An Introduction to WiMAXWiMAX 介紹

暨南大學通訊所

魏學文willncnuedutw

http163222450httpwww80216comncnuedutw

Broadband Local Loop Transmission Lab

暨南大學通訊所

魏學文willncnuedutw

http163222450httpwww80216comncnuedutw

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

WiMAX 是什麼

小學生上課只帶notebook高中生下載一片DVD電影只要十分鐘

大學生跟女朋友用網路電話聊天再久都不怕

研究生可以在校園任何角落上網查資料作研究偶爾偷偷玩線上遊戲

上班族也可以在上班途中用PDA 看電視了

WiMAX生活生活無限寬廣

WiMAX 是什麼

Very high throughput (bit rate)ndash High BWndash High spectral efficiency (high density

modulation)

同時最好

ndash Large coverage (anytime anywhere)ndash Good performancendash Not expensive

IEEE 80216

IEEE（Institute of Electrical and Electronics Engineers電機和電子工程師協會）是世界性的專業組織每年出版許多專業的技術期刊IEEE委員會下設的IEEE 802負責制定網路相關的標準非常有名並廣為業界採用

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

WiMAX 是什麼

小學生上課只帶notebook高中生下載一片DVD電影只要十分鐘

大學生跟女朋友用網路電話聊天再久都不怕

研究生可以在校園任何角落上網查資料作研究偶爾偷偷玩線上遊戲

上班族也可以在上班途中用PDA 看電視了

WiMAX生活生活無限寬廣

WiMAX 是什麼

Very high throughput (bit rate)ndash High BWndash High spectral efficiency (high density

modulation)

同時最好

ndash Large coverage (anytime anywhere)ndash Good performancendash Not expensive

IEEE 80216

IEEE（Institute of Electrical and Electronics Engineers電機和電子工程師協會）是世界性的專業組織每年出版許多專業的技術期刊IEEE委員會下設的IEEE 802負責制定網路相關的標準非常有名並廣為業界採用

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX 是什麼

小學生上課只帶notebook高中生下載一片DVD電影只要十分鐘

大學生跟女朋友用網路電話聊天再久都不怕

研究生可以在校園任何角落上網查資料作研究偶爾偷偷玩線上遊戲

上班族也可以在上班途中用PDA 看電視了

WiMAX生活生活無限寬廣

WiMAX 是什麼

Very high throughput (bit rate)ndash High BWndash High spectral efficiency (high density

modulation)

同時最好

ndash Large coverage (anytime anywhere)ndash Good performancendash Not expensive

IEEE 80216

IEEE（Institute of Electrical and Electronics Engineers電機和電子工程師協會）是世界性的專業組織每年出版許多專業的技術期刊IEEE委員會下設的IEEE 802負責制定網路相關的標準非常有名並廣為業界採用

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX 是什麼

Very high throughput (bit rate)ndash High BWndash High spectral efficiency (high density

modulation)

同時最好

ndash Large coverage (anytime anywhere)ndash Good performancendash Not expensive

IEEE 80216

IEEE（Institute of Electrical and Electronics Engineers電機和電子工程師協會）是世界性的專業組織每年出版許多專業的技術期刊IEEE委員會下設的IEEE 802負責制定網路相關的標準非常有名並廣為業界採用

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IEEE 80216

IEEE（Institute of Electrical and Electronics Engineers電機和電子工程師協會）是世界性的專業組織每年出版許多專業的技術期刊IEEE委員會下設的IEEE 802負責制定網路相關的標準非常有名並廣為業界採用

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IEEE 8021 Higher layer LAN protocols IEEE 8022 Logical link control IEEE 8023 Ethernet IEEE 8024 Token bus (disbanded) IEEE 8025 Token Ring IEEE 8026 Metropolitan Area Networks (disbanded) IEEE 8027 Broadband LAN using Coaxial Cable (disbanded) IEEE 8028 Fiber Optic TAG (disbanded) IEEE 8029 Integrated Services LAN (disbanded) IEEE 80210 Interoperable LAN Security (disbanded) IEEE 80211 Wireless LAN (Wi-Fi certification) IEEE 80212 demand priority IEEE 80213 (not used) IEEE 80214 Cable modems (disbanded) IEEE 80215 Wireless PAN

ndash IEEE 802151 (Bluetooth certification) ndash IEEE 802154 (ZigBee certification)

IEEE 80216 Broadband Wireless Access (WiMAX certification) ndash IEEE 80216e (Mobile) Broadband Wireless Access

IEEE 80217 Resilient packet ring IEEE 80218 Radio Regulatory TAG IEEE 80219 Coexistence TAG IEEE 80220 Mobile Broadband Wireless Access IEEE 80221 Media Independent Handoff IEEE 80222 Wireless Regional Area Network

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Initially 80216 is only a BWAtechnology

High speed connectionPoint to multipoint system Uses radio waves

Source Thikriat Al mosawiSourcehttpwwwmvtcothimagesuploadbig661jpg

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Evolution of IEEE 80216 1999 to presentPoint-to-multipoint broadband wireless accessndash Originally operates in 10-66 GHz spectrumndash Data rates up to 134 Mbpsndash Requires directional line-of-sight (LOS) propagationndash QAM

80216a adopted to address these concernsndash Operates in 2-11 GHz spectrumndash Eliminates need for directional LOS propagationndash Greater range but lower data ratesndash OFDM and OFDMA

80216 has the amendment integrated into it and is called the 80216-2004standard80216e has been finalized at Sep 2005 in Taiwan and is referred to as 80216e-2005 (compatible to 80216-2004)80216 Relay is the next key pointMost likely first implementation will be in the 24GHz ISM and the 51 to 58 GHz U-NII (Unlicensed National Information Infrastructure) band 35 23 27 17 and 19 GHz Licensed bands may be used for TV and VoP

Jim Carlson CEO Carlson Wireless

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX Forum WiMAX的全名是Worldwide Interoperability for Microwave Access一般中文翻譯為「微波存取全球互通」non-profit organization It was formed in 2003 It supports the IEEE 80216 Broadband Wireless Access It has more than 110 340 350+ members such as Alcatel ATampT Intel Nortel Motorola Samsung Siemens Nokia and so forth

WiMAX認證之於80216就好像Wi-Fi認證之於80211

Source Thikriat Al mosawi

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX 技術 80216 技術

We should say (個人看法)80216 技術

WiMAX 產品

WiMAX 認證技術

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX Applications

Multi-player interactive games VOIPVideo conference Stream Video Web Browsing Media contents download

Killer application80216 is the only one carrier

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80216 system

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Intelreg PROWireless 5116Broadband Interface

Highly integrated SoC based on IEEE 80216-2004 standard256 OFDM PHY with support for channel bandwidths up to 10 MHzTDD and HFDD duplexing modesConcatenated Reed-Solomon and Convolutional Encoding (Forward Error Correction)Adaptive modulation (BPSK QPSK QAM16 QAM64)Enhanced link budget supportndash Receive space time codingndash Uplink sub-channelizationndash SNR RSSI channel quality

measurementndash ARQ capable

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Why Taiwan Promotes the 80216 technology

We smell the dollars

Next step of WLANEssential IPNetwork technologyKey step to the 4G

WLAN系統產品 2003年產量達4599萬佔全球91 產值達504億新台幣佔全球42

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Evolution of Mobile Communications1G AMPS2G GSM3G WCDMACDMA2000TD-SCDMAndash 35ndash 39

4G OFDM

IEEE 82011bagn (Data Com)

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

台灣無線通訊產業技術發展理念

附加價值

產業價值鏈

創新研發中心

產品及服務中心

全球營運總部

制定標準

創新

設計

研發

製造

裝配

物流

品牌

服務

行銷

提高產品附加價值

附加價值高

替代性低

台灣科技產業主力推移

技術規劃 核心晶片

台灣廠商新創事業

電信國家型計畫

B3G4GB3G4G 3G3G

(掌握核心晶片)

開拓市場

Copy from ICL

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

M-Taiwan VisionAny where any time any one to enjoy

BroadbandWireless services

M-lifestyle

e-Zoo

ITS

e-Traffic e-Logistics

e-govWireless access + M- applications

WLANWiMAX Cellular M-services

M-learning

Food Guide

Tour Guide

Art Museum

School

Library Medicine

bullFTTHbullxDSL

Copy from NTPO

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

bullGov ServicebullSurveillancebullm-Traffic Servicebullm-Medicarebullhellip

M-Taiwan A Program to Realize TW-WiMAX Blueprint

BroadbandPipeline

FTTHCable

Backbone

Cellular（ GSMGPRS3G

PHS）

Taichun MetroBackbone

Kaohsing MetroBackbone

Taipei MetroBackbone

Access

Netw

ork

AP

WLANWiMAX(Wireless

Broadband）

Dual Network

bullIPTVbullVoIPbullVideo PhonebullHomecarebullhellip

bullCampus SafetybullDistant Learninbullhellip

Broadband Pipeline Mobile Applications and WiMAXWLAN-Cellular Dual Network 1 Billion $USD 220 Million $USD

M-Service

M-Learning

M-Life

AP

AP

Copy from NTPO

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Wireless Taipei City

Schedule Tendered RFP in May 2004 The network infrastructure is now under construction

Business Model

Signed a 9-year BO (Build-Operate) contract with Qware System in Sept 2004 to design construct manage and maintain this wireless network and provide service

Applications VoIP multimedia service SMS remote security system online learning

Population amp Coverage

26 million residents 272 km2(105 square miles)

DeploymentCost

$ 90 millions (USD) for the whole network of10000 access points (expected) It had deployed 5000 AP to provide broadband wireless related access so far

Technology Wi-Fi access with WiMAX backhaul data transmission speed exceeding 05 Mbps per user

The largest Metro-WiFiWiMAX City around the worldThe largest Metro-WiFiWiMAX City around the world

SourceIEKITRI (200412)

Copy from NTPO

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

經濟部WiMAX 加速計畫

Chip setndash MediaTek

BSndash Gemtek ALPHA ZyXEL TECOM COMPAL

CPEndash Gemtek ALPHA ZyXEL CAMEO CyberTAN

Accton MiTAC MW SIndash Vibo ChungHua Telecom TaiwanMobile

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IEEE 80216試驗網路建置與效能評估Establishment and performance evaluation of IEEE 80216

trial network

電信國家型建置計畫參與學校與單位國立暨南國際大學

國立中央大學

中華電信研究所

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

SS 水里商工

4SS暨大校園

SS暨大附中

2BS科一館 SS中壢高中

3SS工二館

2SS研二館

2BS志希館

WLAN

80216-based Mobile Device

中央大學80216網路

暨南國際大學80216網路

TANETTWAREN

AAA

WLAN

SS中大鹿林山天文台 80211 AP

80211 AP

80211 AP

AAA

80216-based Mobile Device

移動式5SS

移動式4SS

SS中華電信研究所

SS工五館

暨南國際大學與中央大學網路通信方式示意圖

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

CHT-TL中壢80216子網路CHT-TL台北

80216子網路

SS

SS

BS

SS

SS

SS

BS

WLAN

SS

移動式

B棟大樓

F棟大樓

E棟大樓

電話大樓行通大樓

總公司大樓

北分大樓

移動式

SS 中央大學

SS

SS

H棟大樓

專線

數分大樓

SS

中華電信研究所中壢台北園區網路通信方式示意圖

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

已穫得35GHz實驗頻段

TDDndash 3414MHz至3435MHz (21MHz)

FDDndash Uplink 3400MHz至3428MHz (14MHz)ndash Downlink 3500MHz至3528MHz (14MHz)

TL中壢台北園區網路FDDNCNU amp NCUFDD amp TDD

感謝電信總局支持本計畫

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

58GHz WiMAX Trial Network

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Pre-16e Network

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

系統建置照片

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

測試路線圖(二) - 科技學院

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

量測結果(四)科院前道路(40Kmh)

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

6km

地理中心碑

BS

埔里定點測試(一)地理中心碑

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

埔里定點測試(一)地理中心碑

Uplink RSSI (dBm) -843

Uplink SNR (dB) 2120

Uplink Current Rate QAM64 34

Downlink RSSI (dBm) -77

Downlink SNR (dB) 28

Downlink Current Rate BPSK 12

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

開放實驗室

中大和暨大已各自成立一個開放實驗室透過開放實驗室國內學校及研究單位使用者可以直接access 此網路並利用此80216 試驗網路作為相關研究計畫的驗證平台

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80216建置計畫網頁http163221963 httpwww80216comncnuedutw

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

展示項目

80216 WebCamWiMAX連線效能量測

VoIP over WiMAXWiMAX 網路之影像電話展示

即時傳訊服務測 試

多媒體應用教學服務

IPTV over WiMAX

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80216 WebCam

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX連線效能量測

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

VoIP over WiMAX-CO

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

VoIP over WiMAX-RT

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX 網路之影像電話展示-CO

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX 網路之影像電話展示-RT

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

多媒體應用教學服務

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IPTV over WiMAX

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

建置平台可提供之服務

各種環境場測（Field trials）

提供定點測試校園移動測試跨網測試環境

提供新服務驗證測試

80216 網管研究平台

提供傳收機測試提供一定點侵入式（可更動硬體）測試連結

研究計畫平台支援

Others

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

條條道路通羅馬

高速

中速

低速

移動性

資料傳輸率

144 kbps 144 kbps 384 kbps lt50 Mbps lt100 Mbps

1G(類比)

2G(數位)

3G(IMT2000)

3G+

AMPSETACSJTACSNMT

WLAN

High speedWLAN

80216WiBro

WPAN

CDMAGSMTDMACDMA2000 EV-DODV

W-CDMAHSDPA

1995 2000 2005 2010+

80211abg

BluetoothZigbee

80211n

語音 資料影像 視訊多媒體

4G

WiMAXWiBro

TW4G-Mobil

資料來源 Samsung

LTE(IMT-Advanced)

IEEE80216j

Copy from ICL

16m

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IMT-Advanceddoc IEEE 80211-070375r0

ITU = International Telecommunication Unionndash ITU-R = Radio communication Sector

bull SG 8 = Study Group 8 ndash Mobile Radiondash WP 8F = Working Party 8F ndash IMT-2000 amp IMT-Advanced (第17次會

議上ITU給了B3G技術一個正式的名稱IMT－Advanced )

IMT-Advanced allows for two new radio interfaces mobile access and nomadic local area access ndash Mobile 100 Mbps for high mobilityndash Nomadic access 1 Gbps

Deployment after 2010

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

The ITU-R vision for systems beyond 3G

The ITU-R vision for systems beyond 3G comprises two major paths On one hand existing and evolving access systems will be integrated on a packet-based platform to enable cooperation and interworking of these systems in the sense optimally connected anywhere anytime On the other hand the radio access system for new mobile access and new nomadiclocal area wireless access will be developed to provide access with significantly improved performance compared to todays systems The focus of the WINNER project is the development of this radio access system by taking into account the interworking with other systems

ndash The envisioned capabilities of the new components of future mobile and wireless communication systems were agreed with the following peak aggregate user data rates

ndash up to approximately 100 Mbps for the new mobile access and up to approximately 1 Gbps for new nomadic local area wireless access

Recommendation ITU-R M1645

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Bit Rate Target in 4Ggt 100 Mbps for mobile 250kmhr-350kmhrgt 1 Gbps for time-invariant environment (in hot spots amp indoor)

BW 20MHz (5-100MHz)Spectral efficiency 5-20 bpsHz

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

DoCoMo 4G field experiment 在日本神奈川縣橫須賀進行的現場試驗中Docomo研究人員使用100MHz頻寬每小時10公里的速度運動的行動站進行的下行連結傳輸速率達到

5Gbps (Dec 25 2006)使用VSF-Spared OFDM (Variable Spreading Factor -Spread Orthogonal Frequency Division Multiplexing)MIMO 12x12頻譜效率50bpsHz (5Gbps100MHz)

httpchinanikkeibpcojpchinanewsnewsmobi200702140119html

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Turbo coding (R=13-56 18116)Max-Log-MAP decoding

Channel codingdecoding

QPSK 16QAM 64QAMQPSK 16QAM 64QAMData Modulation

UDPDCH 4816 UDPCCH 64RACH 16

DSPDCH(DDPDCH) Max 128 (2-D spreading time domain max 16)

Spreading Factor

16384 McpsChip rate

7585 ms + GI 1674msOFCDM symbol duration

2768Number of sub-carriers

40MHz100MHzBandwidth49 GHz4635 GHzCarrier Freq

UplinkDownlink

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

B3G in ChinaFuTURE (2001~200308)ndash 7 universities proposals accepted (東南清華北郵等)

FuTURE+ (200308~200512) 15億RMBndash FDD group WG (東南 lead)ndash TDD group WG (北郵 lead)ndash Unified HW platform WGndash Simulation and modeling WG (channel QoS traffic resource)

FuTURE II (2006~2010)ndash 863四大專項之一 預估每年2億RMB

Combination of GMC amp OFDMndash Turbo Receiver Distributed MIMO etcndash Target Spectrum Efficiency 5bpsHz ie 100Mbps in 20MHz bandndash Frequency band 35GHz now 51GHz later

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

TW4G台灣第四代行動通訊

Form a dedicate 4G team for long-term RampD work Focus on IPR creation contributions to the standard bodies and publications

TW4G

TW4G網址httpwwwtw4gntpoorgtw

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

4G or neverWhere is the Band Penetration capability80216-2004 or 80216e-2005 or 16mCOST carrier grade network or hot spot only後起之秀

ndash 80220 80222ndash 3G LTE

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Source Gartner Research

Hype Cycle for New Technologies

Hype Cycle Stages

1 Technology Trigger

2 Peak of Inflated Expectations

3 Trough of Disillusionment

4 Slope of Enlightenment

5 Plateau of Productivity

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Source Gartner Research

Some cases

80216

80220

4G

3G

802112G

LMDS

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80220的成立Mobile-Fi

IEEE 80220 工作小組在2002年12月成立也被稱為「行動寬頻無線存取小組（Mobile Broadband Wireless AccessMBWA）」

參與的公司有ArrayCom Cisco Flarion HP-Compaq Lucent Motorola Nokia Qualcomm Navini Nextel Texas Instruments and Samsung

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Source IEEE P 80220trade V14

80220 FeaturesCharacteristic Target Value

Mobility Vehicular mobility classes up to 250 kmhr (as defined in ITU-R M1034-1)

Sustained spectral efficiency gt 1 bsHzcell

Peak user data rate (Downlink (DL)) gt 1 Mbps

Peak user data rate (Uplink (UL)) gt 300 kbps

Peak aggregate data rate per cell (DL) gt 4 Mbps

Peak aggregate data rate per cell (UL) gt 800 kbps

Airlink MAC frame RTT lt 10 ms

Bandwidth eg 125 MHz 5 MHz

Cell Sizes Appropriate for ubiquitous metropolitan area networks and capable of reusing existing infrastructure

Spectrum (Maximum operating frequency) lt 35 GHz

Spectrum (Frequency Arrangements) Supports FDD (Frequency Division Duplexing) and TDD (Time Division Duplexing) frequency arrangements

Spectrum Allocations Licensed spectrum allocated to the mobile service

Security Support AES (Advanced Encryption Standard)

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80220 Vs 80216

80216 Based on existing standardsndash 80220 Start from scratch

80216 Emphasizes on throughput rather than mobilityndash 80220 Developed to cover the mobility part of

8021680216e Speeds up to vehicular speedsndash 80220 Speeds upto 250Kmhr

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80222

In October 2004 IEEE set up a working group to develop the 80222 Standard for Wireless Regional Area Networks (WRAN)The idea behind 80222 is that there are considerable unused frequencies between VHF and UHF broadcast channels between 54 and 865 MHz - which could be used to beam wireless broadband as far as 40 kilometers to serve areas not well served by alternatives such as cable or DSLCognitive Radio（感知無線電）

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

3GPP Long Term Evolution (LTE)

LTE

LTE

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Key tech in the HSDPA

Adaptive Modulation and CodingAdaptive SchedulingHybrid ARQDownlink shared channel

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Key tech in the LTE

OFDMA for the downlinkSC-FDMA (DFT-spreading OFDM) for the uplinkndash Low PAPRndash CPndash Localized or distributed

transmission

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

M-pointDFT

Spreading

Symbol to subcarrier mapping

N-pointIFFT

0

0

0

0

Localized contiguous subcarriers

Distributed evenly spaced subcarriers

C80220-05-90

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OUTLINESWhat is WiMAXWhy we promote WiMAX in TaiwanWiMAX in TaiwanWiMAX 4G or neverCore technologies and research topics in WiMAX

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Core Technologies in 80216

OFDMAAdaptive ModulationSync and CFOFEC code RS PCC BTC CTC LDPCMIMO Multi-hop relayMobility Supporting in 16e and 16m others

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Multiple access in OFDMOFDM + TDMAOFDM + FDMA = OFDMAOFDM + CDMA = MC-CDMA

OFDMA是一個多重接取的技術可根據通道狀況改變系統頻寬與子載波數運作原則是將所有子載波分割成若干群組稱為子通道 (sub-channel) 分配給不同用戶使用並根據傳輸環境狀況決定各個子通道的子載波數另外藉由子載波配置與適應性調變和編碼 (AMC) 技術可讓OFDMA之實體層方便在通道變化較大的移動環境中傳送資料

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Broadband Local Loop Transmission Lab

Frequency domain description Null carrier

User 1

User 2 User 3

--Data subcarriersfor data transmission

--Pilot subcarriersfor various estimation purposes

--Null carrierno transmission at all for guard band and DCcarrier

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

The FCH contains the DL_Frame Prefix and specifies the length of the DL-MAPmessage that immediately follows the DL_Frame _Prefix and the repetition codingused for the DL-MAP message

contains the DL_FramePrefix

use therepetition code

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Sub-channel (PUSC)

2048 case

24602048=075

2 pilots

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Subchannel allocation in the downlink may be performed in the following ways

partial usage of subchannels (PUSC) some of the subchannels are allocated to theTransmitter

full usage of the subchannels (FUSC) all subchannels are allocated to the transmitter

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Sub-channel (FUSC)

1024

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Tile and sub-channel for uplink

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

OFDMA SLOTFor downlink FUSC and downlink optional FUSC using the distributed subcarrier permutation one slot is one subchannel by one OFDMA symbolFor downlink PUSC using the distributed subcarrierpermutation one slot is one subchannel by two OFDMA symbolsFor uplink PUSC using either of the distributed subcarrier permutations and for downlink TUSC1 and TUSC2 one slot is one subchannel by three OFDMA symbolsFor uplink and downlink using the adjacent subcarrierpermutation one slot is one subchannel by one two three or six OFDMA symbols

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Data Region

In OFDMA a Data Region is a two-dimensional allocation of a group of contiguous subchannels in a group of contiguous OFDMA symbols All the allocations refer to logical subchannels This two dimensional allocation may be visualized as a rectangle such as the 4 times 3 rectangle shown in Figure 215

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Slot and Data Region

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

-- Group 0 includes cluster 0~23 288 (=2412)Group 1 includes cluster 24~39 192 (=1612)Group 2 includes cluster 40~39 288 (=2412)Group 3 includes cluster 64~79 192 (=1612)Group 4 includes cluster 80~103 288 (=2412)Group 5 includes cluster 104~119 192 (=1612)

Subcarriers (erasing pilots)

PermutationBase 12

PermutationBase 12

PermutationBase 12

PermutationBase 8

PermutationBase 8

PermutationBase 8

Broadband Local Loop Transmission Lab

Example 2048 OFDMA PUSC

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Step1 The carriers for subchannel s=1 in IDcell=0ndash Nsubchannels=12ndash Nsubcarrier=24ndash k=0 1 hellip23 s=1(user2)ndash permutation sequence12=69481011527310

Step2 nk mod Nsubcarrier = (k+13s) mod Nsubcarrier = (13 14 hellip 36) mod 24= (13 14 15 16 17 18 19 20 21 22 23 0 1 2 3 4 5 6 7 8 9 10 1112)nk mod Nsubchannels = 12hellip11012hellip110

Step3 ps[nk mod Nsubcarrier ] = 9481011527310694810115273106ps[1 ] = 4 ps[2] = 8 hellip

Step4 Nsubchannels nk =156 168180hellip 276 0 1224 hellip 144Step5 Nsubchannels nk + ps[nk mod Nsubcarrier ]

=160 1761902032092182352432532642829163246596574 9199109120138153

Example for Group 0

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

Use permutation to control the sub-carrier allocation pattern

It can be interleaved or sub-band

80216-2004 OFDMA - PermutationBase 12

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

bullbullbullbullbullbull

Broadband Local Loop Transmission Lab

80216-2004 OFDMA - PermutationBase 8

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Logical tiles are mapped to physical tiles( ) ( [( ) mod ] _ ) mod

( ) the physical tile index in the FFT with tiles being ordered consecutively from the mostnegative to the most positive usedsub

subchannels subchannels subchannelsTiles s n N n Pt s n N UL PermBase Nwhere

Tiles s n

= sdot + + +

subchannels

carrier(0is thestarting tile index) the tile index 0amp5 in a subchannel the tile permutation the subchannel number in the range 0ampN -1

_ an integer value in the range 0amp69 which is

nPtsUL PermBase assigned by a management entity

the number of subchannels for the FFT sizesubchannelsN

The mapping of data onto the subcarriers( ) ( 13 ) mod

( ) the permutated subcarrier index corresponding to data subcarrier n is subchannels a running index 0amp47 indicating the data constellation p

subcarrierssubcarrier n s n s Nwhere

subcarrier n sn

= + sdot

oint the subchannel number the number of subcarriers per slotsubcarriers

sN

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

ExampleStep1 The tiles used for subchannels = 3 in UL_PermBase = 2

bull Number of subchannels Nsubchannels = 70bull Number of subcarriers in each OFDMA symbol= 24bull Number of data subcarriers in each subchannel Nsubcarriers = 48bull TilePermutation = 6 48 58 57 50 1 13 26 46 44 30 3 27 53 22

18 61 7 55 36 45 37 5215 40 2 20 4 34 3110 5 41 9 69 63 21 11 12 19 68 56 43 23 2539 66 42 16 47 518 62 14 33 24 32 17 54 2967 49 65 35 38 59 64 28 60 0

Step2 Apply the permutation due to the selection of the subchannel(s) rotate three times 57 50 113 26 46 44 30 3 27 53 22 18 61 7

55 3645 37 52 15 40 2 20 4 34 31 10 5 41969 63 21 1112 19 68 56 43 23 25 39 66 4216 47 51 8 62 14 33 24 32 17 5429 67 49 65 35 38 59 64 28 60 0 6 4858

Step3 Take the first six numbers add the UL_PermBase (perform modulooperation if needed) 59 52 3 15 28 48

Step4 Add the appropriate shift 59 122 143 225 308 398

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Example17 usersrsquo sub-carrier allocation for 512-ofdma uplink system

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER2 USER2 USER2

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER9 USER9 USER9

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER11 USER11 USER11

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER17 USER17 USER17

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER4 USER4 USER4

USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER14 USER14 USER14USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER10 USER10 USER10

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER12 USER12 USER12

USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER15 USER15 USER15USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16USER16 USER16 USER16

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Advantages of OFDMAFrequency reuse (reuse factor of 1 is possible max sectors allocation)Adaptive carrier allocations (will be very powerful to combine with AMC) (work for different users) (timefreq diversity)Larger coverage and penetration

source IEEE80222-05-0005r1

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Adaptive Modulation

Maximize throughputMinimize average powerMinimize average BER

Throughput Coverage

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Adaptive Modulation

Change modulation according to channel quality ndash Maximize throughputndash Minimize average BER

Changendash Constellationndash Transmit powerndash Coding scheme

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Adaptive Modulation and CodingModulation Block Size Over Coging Rate RS CC

BPSK 12 12 (16122) 23

QPSK 24 12 (32244) 23

QPSK 36 34 (40362) 56

16-QAM 48 12 (64488) 23

16-QAM 72 34 (80724) 56

64-QAM 96 34 (108966) 34

64-QAM 108 56 (1201086) 56

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

瞬時的BER曲線

0 5 10 15 20 2510-7

10-6

10-5

10-4

10-3

10-2

10-1

1008-Mode Modulation Over AWGN

ES N0

BE

RQPSK + CC (23) + RS (32244)QPSK + CC (56) + RS (40362)16-QAM + CC (23) +RS (64488)16-QAM + CC (56) + RS (80724)64-QAM + CC (34) + RS (108966)64-QAM + CC (56) + RS (1201086)BPSK + CC (23) + RS (16122)

Design BERDesign BER

SS11 SS22 SS33 SS44 SS55 SS66 SS77

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

How to Choose the Switching Levels

There were many methods for determining the switching levels such as Limiting the Peak Instantaneous BER Torrancersquos method and Lagrangian method Instantaneous BER is a very simple way

10 15 20 25

10-4

10-3

10-2

10-1

100

Average channel SNR (dB)

BE

R

design Pth = 10 -3 fdT = 00001

Lagrangian methodInstantaneous BER method

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Synchronization

Network SyncFrame syncSymbol SyncCFOIQ ImbalancePhase Noise

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Why Carrier Frequency Offset Exist

The Local Oscillator Tx amp Rx do not have the same frequencyDoppler Effect

x(t) X

2 cj f te π

X

ˆ ˆ(2 )cj f te π θminus +

(2 )( ) ( ) j ftr t x t e π θΔ +=

(2 )( ) ( ) cj f tr t x t e π θ+=

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Effects of a Carrier frequency Offset

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 2

Sub_Ch = 64 CP Ratio = 18QAM = 4 points

Offset = 1

Integer CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 02

Fractional CFO

Sub_Ch = 64 CP Ratio = frac14QAM = 4 points

Offset = 01

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IQ imbalancebull省略掉將RF降至IF所需的IF filter 這樣的做法稱 ldquoDirect conversionrdquo or ldquoZERO-IFrdquobull因為直接從RF降至baseband所以在mixer中會出現 In-phase 跟 Quadrature之間有gain 跟 phase 的 imbalance 的現象

)sin()1()()cos()(

ϑε ++==

twtQtwtI

c

c

Gain imbalance Phase imbalancelt90

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IQ imbalance effect

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

IQ imbalance effect (in multipath channel)

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Effects of Oscillator Phase Noise

From httpeesoftmagilentcompdfwireless_networking_04pdf

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Lorentzian ModelA practical oscillator does not produce a carrier at exactly one frequencybut rather a carrier that is phase modulated by random phase jitter As aresult the frequency is never perfectly constant thereby causing ICI

From httpeesoftmagilentcompdfwireless_networking_04pdf

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

f1 f2 f3 fc+f1 fc+f2 fc+f3

fc

fc f1 f4f3f2 fc+f1 fc+f2 fc+f3 fc+f4

1) N = 4

2) N = 3

ICI when N=4

ICI when N=3

Phase noise effects on OFDM(contd)

httpwwweceutexasedu~wirelessEE381K11_Spring03projects14ppt

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Effect of Oscillator Phase Noise ( Contrsquod )

From httpeesoftmagilentcompdfwireless_networking_04pdf

Given a fixed bandwidth the greater the number of sub-carriers more susceptible is the overall system to phase noise

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

MIMO-OFDM

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

MIMO (Mulitple-input multiple output)

helliphellip

helliphellip

TX RX

( )tH k τ

)(ty)(tx

( )th k 21 τ ( )th k 22 τ

( )th k 12 τ

( )th kMM TR τ

( )th kM R1 τ

( )th k 11 τ

1

2

RMTM

2

1 ( )th kMT1 τ

( )th kM R2 τ ( )th kMT

2 τ

bull Increase the capacity or improve the performance

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

2 by 2 Example

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

MIMO R2T2 vs R1T1

Double capacity or improve the performance

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Combine MIMO and OFDM

OFDM-MOD

OFDM-MOD

OFDM-DMOD

OFDM-DMOD

MIMO Channel Model )( fR)( fS

)( tfH qp

1

TM RM

1

When will be MIMO 好吃又不貴

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

COST207 TU

MIMO-OFDM R2T2

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Space Time Code in MIMO

Space-Timendash 結合了channel coder之設計與多根傳送天線之應用

Bell Labs Layered Space-Time (BLAST)Space-Time Trellis Code (STTC)Space-Time Block Code (STBC)Space-Frequency Block Code (SFBC)

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Review of STBC

STBCndash 1998 ATampT Alamoutindash Tx

bull Mapping operation of a block of input symbols into space and time domains

bull Creating orthogonal sequencesndash Rx

bull Channel estimationbull Combining procedurebull Maximum Likelihood (ML) detection rule

ndash A very limited coding gain is expected

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

STBC-OFDM系統

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+⎥⎦

⎤⎢⎣

⎡

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

minus

minus=

⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢

⎣

⎡

+

+

+

+

+

+

++

++

++

++

+

+

+

+

)2(12

)2(02

)1(12

)1(02

12

02

)12(12

)22(12

)22(02

)12(02

)11(12

)21(12

)21(02

)11(02

)2(12

)2(02

)1(12

)1(02

n

n

n

n

n

n

nn

nn

nn

nn

n

n

n

n

WWWW

XX

HHHHHH

HH

YYYY

OIST

OIST

OIST

OIST

22222222 WXHY +=

2I2O STBC-OFDM

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

STC using 2 antennas in PUSCThe transmission of the data shall be performed in pairs of symbols as illustrated

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

2I2O STBC-OFDM之效能

0 5 10 15 2010-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

EbN0(dB)

Bit

Erro

r Rat

e

Full v=250Full v=125Full v=75LS-ZF v=250LS-ZF v=125LS-ZF v=75

COST207 TU 8 pathsCOST207 TU 8 paths

消除消除ICIICI並獲得接收分並獲得接收分集增益集增益

ICIICI破壞信號正交性破壞信號正交性

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Adaptive Antenna System1根據用戶端有不同的空間傳播方向會提供不同空間通道特性Adaptive Antenna運用數位訊號處理和陣列天線觀念充分利用訊號方向性來計算權重並適應性的調整權重

2運用波束(Beam)行成技術來控制波束場型(Beam Pattern)的調整把主波束對準目標訊號並適應性地即時追蹤訊號同時抑制干擾訊號以強化接收品質並增加容量擴大涵蓋面和提高傳輸速率

3Adaptive Antenna 的效益雖然高但也因需具備高複雜度的適應性演算法使得硬體實現上的難度增高為滿足無線通訊高頻譜效率需求目前Adaptive Antenna是Smart Antenna運用的主要類型一般採用天線陣列數目為4~12個天線單元

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Switched Beam and Adaptive Antenna

Capacity or performance

interference

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

MIMOsmart antenna and multi-hop network

Constantmax throughput over whole cellEnlarge diversity orderLess hops

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

FEC codes in 80216Block codendashReed-Solomon（RS） codendashTurbo product codendashLDPC (low density parity check) codeConvolutional code ndashPuncture convolutional codendash Convolutional turbo code (CTC)

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Block code

Well-developed theoryt =(dmin -1)2 dmin =codewords 間最小距離

encoding

Parity-check bitsSystemaic form

Blocking message stream

codeword

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Principle of block error correction

Codeword 1 Codeword 2

1 2 3 4 5 6 6 5 4 3 2 1

dmin = 13 case t=6

Received word

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Linear Block code(encodingdecoding)

C1timesn = u1timesk Gktimesn Gktimesn is the generator matrixAll codewords C span the null space of a parity check matrix H (eg C1timesn HT

(n-k)timesn =0)GHT=0Denoted as (n k) code

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphk-10 hk-11 hellip hk-1n-1

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Reed-Solomon (RS) codePresent in 1960 Cyclic code a special case of non-binary BCH codeGF(28) is the most popular employed fieldRandom and burst error correcting capabilityParity check = 2t = dmin-1 (MDS code)High code rate

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

RS code in 80216

(255239) RS over GF(28) t=8Shorten RS code (239-krsquo)Punctured parity check symbols for variable error correcting capability Punctured parity check symbols can be regarded as erasures in the decoding

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

RS code(t=8)在64QAM

4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS code performance over 64-QAM

EbNo(dB)

Pb(

E)

theorem RS(204188) shaping+64QAM

BER=10-3時編碼增益為15dBBER=10-4時編碼增益為27dB

Coding Gain

coding gain is the measurement in the SNR difference between the uncoded and coded system required to reach the same BER

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Convolutional code

][][0

knxwnyN

kk minus=sum

=

DD - - -D DD

w1 w2 w3 - - - wN-1 wNw0

SUM

x[n]

y[n]

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Convolutional codePRESENTED BY ELIAS IN 1955BASED ON LINEAR CONVOLUTION THM

bullRSC CAN HAVE THE SAME FREE DISTANCE

bullRSC HAVE BETTER PERFORMANCE AT LOW SNR

bullRSC HAVE BETTER PERFORMANCE WHEN PNUCTURED

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

PUNCTURED CONVOLUTIONAL CODEIMPROVE THE CODE RATE

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Soft decision and hard decision

-2 0 2 4 6 8 10 12 14-5

-45

-4

-35

-3

-25

-2

-15

-1

-05Punctured cnvolutional code(12)hard decis ionampdemapping16QAM

EbNo(dB)

Pb(

E)

theorem shaping hard decis ionampdemappingsoft decis ionampdemapping

在BER=10-3時soft hard 編碼增益相差25dB

8-9 dB

VITERBI ALGORITHM (1967)- MAXIMUM LIKELIHOOD DECODING (FORNEY 1973)- SOFT-DECISION DECODING FLOATING OR FIX-POINT- HARD-DECISION DECODING

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Concatenated code

Outer EncoderRS code Interleaver Inner Encoder

frac12 CC code

Outer DecoderRS code De-interleaver Inner Decoder

frac12 CC code

Channel

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

RS code+Punctured Convolutional code

2 4 6 8 10 12 14 16 18-45

-4

-35

-3

-25

-2

-15

-1

-05RS(204188)+punctured convolutional code (12) over 64QAM

EbNo(dB)

Pb(

E)

theorem shaping convolutional code(12) RS+Convolutional RS(204188)

在BER=10-3時編碼增益有99dB

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Trubo Code (convolutional turbo code)

Presented by C Berrou A Glavieux and P Thitimajshima in 1993（發表於ICCrsquo93目前引用

次數3505）

parallel concatenation of two RSC codes separated by an interleaver (permutation)

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correcting coding and decoding Turbo-codesrdquoProc ICCrsquo93 Geneva Switzerland May 1993 pp1064-1070

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Trubo decoding means iterative

Code 1 Code 2

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

C Berrou A Glavieux and P Thitimajshima ldquoNear Shannon limit error-correctingcoding and decoding Turbo-codesrdquo Proc ICCrsquo93 Geneva Switzerland May 1993pp1064-1070

Copy from Catherine Douillard

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Duo-Binary Turbo-codes in 80216

Information bits are encoded by couples [1]

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Advantages of Duo-binaryTurbo-codes [1] [2]

Good performance for a very wide range of blocksizesHighly flexible scheme enabling a very fine granularityndash Same encoderdecoder for all blocksizescoding ratesndash Several trade-off in performance (number of iterations decoding

algorithm) implementation complexity (degrees of parallelism)

Hardware benefitsndash Half as many states in trellisndash Smaller loss due to max-log-MAP decodingndash Reasonable complexity Approximately 35 decrease in complexity

per decoded bit compared to Binary TC

1 IEEE 80222-060005r02 Matthew Valenti Rohit Iyer Seshadri ldquoTurbo and LDPC Codes Implementation

Simulation and Standardizationrdquo West Virginia University June 7 2006

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Performance of CTCDuo-Binary TC 8 iterations Max-Log-MAP decoding

IEEE 80216e structuredLDPC BP decoding 50 iterations

AWGN R=12 QPSK

N=576 and 2304 (codedblocksize)

IEEE 80222-060005r0

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Turbo product code

Product codes first described by Elias (1954)Iterative decoding of product codes described by Tanner (1981) Lin amp Costello (1983) and others [1]Turbo product code by Pyndiah in 1994 [2]-[3]

1 IEEE 80216t-00012 R Pyndiah A Glavieux A Picart and S Jacq Near optimum decoding of product codes in

Proc IEEE Global Telecommun Conf GLOBECOMrsquo94 San Francisco CA Dec 1994 pp 339-343

3 R Pyndiah Near optimum decoding of product codes Block Turbo Codes IEEE Trans Commun vol 46 no 8 Aug 1998 pp 1003-1010

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Product code

Information bits

Checks on columnsChecks onchecks

Checks on rows

n1

n2

k1

k2

Very powerful at lowvery low error rates

Copy from Gian Paolo Calzolari

CP = C1otimesC2

C1(n1 k1 d1) C2(n2 k2 d2)

CP(nP kP dP)

kP = k1middotk2

nP = n1middotn2

dP = dmin= d1middotd2

Special case C1= C2 kP = k2 nP = n2 dmin = d2 RP = k2n2

Extended Hamming Codes CP = (EHl)2 are generally preferred since they permit toincrease the product code minimum distance from dmin = 9 to dmin = 16

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Performance of TPC in 80216

Xilinx Solutions forWiMAXWiBroSystem Design Oct 2005

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Low Density Parity Check Codelinear block code base on very sparse Parity-Check matrix（ small number of ones per column and per row ）Invent by Gallager at 1962 In 1996 D MacKay and R Neal show it to have very good performanceRegular Irregular Codes (RichardsonUrbanke 1998) (08dB gain)

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Linear Block CodeAll codewords x span the null space of a parity check matrix H (eg Hx=0)H is a sparse binary matrix with where the set row and column elements are chosen to satisfy desired weight

h00 h01 hellip h0n-1

h10 h11 hellip h1n-1

hellip hellip hellip helliphm-10 hm-11 hellip hm-1n-1

x0

x1

hellipxn-1

= 0

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Tree-LDPC codes in 80216e

H可由mbtimesnb個子矩陣Hb來描述之每個元素表示為一個二進制基底矩陣(Pij) Pij是一個ztimesz的排列矩陣或是零矩陣m=zmbn=znb而nb固定為24mb及z則會隨著不同的碼率而有所改變

nmnmnmmmm

nn

nn

nn

bbbbbbb

bb

bb

bb

timesminusminusminusminusminusminusminus

minusminus

minusminus

minusminus

⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢

⎣

⎡

=

1121211101

1222221202

1121211101

1020201000

PPPPP

PPPPPPPPPPPPPPP

Η

ΚΚΚΚΚΚΚ

ΚΚΚ

HHb2b2

bull 單位矩陣及零矩陣分別以0和-1表示之bull 大於等於1的值則表示為單位矩陣向右做循環位移次數bull H=[(Hb1)mbtimeskb (Hb2)mbtimesmb]分別相對應資料位元及檢查位元bull 為了使編碼具有systematic特性故意將Hb2排列成低密度的下三角矩陣bull 在每個碼率排列方式的kb+2到nb行都是採用雙對角線排法且第kb+1行會有奇數個排列值第一個排列值及最後一個排列值必定相同這樣的排列是為了避免在Hb1裡有任兩行有相同的排列而出現周長為4的情況

bull 此種排列方法的低密度碼必為系統碼

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80216e LDPC碼規格表

k(bytes) Number of subchannels

R=12 R=23 R=34 QPSK 16QAM 64QAM

576 72 24 36 48 54 6 3 2

672 84 28 42 56 63 7

768 96 32 48 64 72 8 4

864 108 36 54 72 81 9 3

960 120 40 60 80 90 10 5

1056 132 44 66 88 99 11

1152 144 48 72 96 108 12 6 4

1248 156 52 78 104 117 13

1344 168 56 84 112 126 14 7

1440 180 60 90 120 135 15 5

1536 192 64 96 128 144 16 8

1632 204 68 102 136 153 17

1728 216 72 108 144 162 18 9 6

1824 228 76 114 152 171 19

1920 240 80 120 160 180 20 10

2016 252 84 126 168 189 21 7

2112 264 88 132 176 198 22 11

2208 276 92 138 184 207 23

2304 288 96 144 192 216 24 12 8

n(bits) n(bytes) z factor

24times96=2304

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

RU encoding (Linear Time Encoding)Richardson amp Urbanke irregular construction(RU algorithm)use the triangular form describe the parity-check matrix

⎟⎟⎟⎟⎟⎟⎟⎟

⎠

⎞

⎜⎜⎜⎜⎜⎜⎜⎜

⎝

⎛

=⎟⎟⎠

⎞⎜⎜⎝

⎛

100110110100110011001010111000011001011101110000001010001111000101100111

EDCTBA

( )TTT BpAuTp 11

2 += minus

( ) TT uCAETp += minus11

利用高斯消去法轉換成H=[I PT]可以很容易的得到矩陣G=[P I] but複雜度高達O(n3)

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

80216 LDPC80216 LDPC效能模擬效能模擬((解碼解碼3030次次QPSK)QPSK)

05 1 15 2 25 3 35 4 45 5 55 610-6

10-5

10-4

10-3

10-2

10-1

100

EbNo(dB)

BE

RSingle-Carrier system over AWGN channel

LDPC Decoder performance

N=576R=12N=1056R=12N=1824R=12N=2304R=12N=576R=23N=1056R=23N=1824R=23N=2304R=23N=576R=34N=1056R=34N=1824R=34N=2304R=34N=576R=56N=1056R=56N=1824R=56N=2304R=56

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

High Mobility Supporting

Channel Estimation amp Frequency-domain EqualizationDetection

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Some definitions

Sub-carrier spacing = symbol rate of OFDMSymbol period TGuard period Δ Overhead (Δ +T) TDoppler rate fdNormalized Doppler rate = fdT

OFDM is very sensitive to the timing and channel varyingAnti multi-path 能力 Δ uarr TuarrNuarrAnti time varying能力 T darrN darr

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Multi-path fading chanel

12

34

5 010

2030

4050

0

0005

001

0015

002

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Normalized Doppler Frequency fDT

ffDDTT = 1e= 1e--2 2 每每1e2 symbols 1e2 symbols 大約會有一次大約會有一次cyclecycle的變化的變化

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

WiMAX2 (80216m)

WiMAX2 is not WiMAXWill be an IMT-advanced proposal100Mbs for mobile1Gbs for fixednomadic High mobility (350kmhr )OFDMAMIMO5 - 25 bsHz spectrum efficiencycell edge bit rate

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

High mobility 16eCarrier frequency 25 GHz

BW 5 MHz

Power delay profile COST 207 Tapped delay line model

Environment TU BU HT

max delay spread 5μs 10μs 20μs paths 12 12 12 Fading Jake

Resolution of path spacing 15M=02μs Time profile spacing 15M=02μs Sub-carriers (N) 512

Sub-carrier spacing 976 kHz

CP overhead 18 for TU and BU 14 for HT

OFDM symbol duration (Ts) (wo GI)1024μs DL UL

Number of guard subcarriers 91 103

Number of used subcarriers 421 409

Number of data subcarriers 360 272

Number of pilot subcarriers (uses both variable and constant sets)

60 136

OFDMA Number of users

15 17

Normalized Doppler freq 0001 0025 008

Mobility 4 kmh 100kmh 300 kmh

Doppler spread 10Hz 250Hz 1000Hz

Environment Pedestrian Vehicular High speed rail

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

H Matrix

Y=F y=FGFHX +Fn H=FGHF W=Fn=HX +W

YHX 1ˆ minus=信號偵測

H為頻域通道響應矩陣

在非時變通道中H會是diagonal matrix

而在時變通道中會產生ICI

非時變通道可使用one tap FEQ

時變通道中需使用H-1作為

Zero Forcing ICI 消除器的係數

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

5 10 15 20 25 30 35 40 4510-5

10-4

10-3

10-2

10-1

SNR dB

BE

R

one tap feq fdT=008 one tap feq fdT=0008 ICI canceller fdT=008 ICI canceller fdT=0008Theory Rayleigh Limite

one tap equalizer 與 ICI canceller 在fdT=008 and fdT=0008下的比較

One tap FEQ vs ICI canceller

在通道變化快速時使用

ICI canceller 是必要的

要得到 ICI caceller的係數

必須要估測通道的變化

Thanks and QA

Thanks and QA