Transceiver Architectures

김 영 석김 영 석

충북대학교 전자정보대학

2010.9.1.9.

Email: kimys@cbu.ac.kr

전자정보대학 김영석 3-1

Contents3. Modulation and Detection

3.1 General Considerations

3.2 Analog Modulation

AM PM FM• AM, PM, FM

3.3 Digital Modulation

• BPSK, BFSK, QPSKBPSK, BFSK, QPSK

5. Transceiver Architectures

5.2 Rx Architectures

• Heterodyne, Homodyne, Image-Reject, Digital-IF, Subsampling

5.3 Tx Architectures

Di t C i T t• Direct-Conversion, Two-step

전자정보대학 김영석 3-2

3.1 General Considerations (Modulation)Definitions

( ) B b d Si l: O h t i i th(a) Baseband Signal: One whose spectrum is non-zero in the vicinity of w=0 and negligible elsewhere

(b) Passband Signal: one whose spectrum is non-zero in a band paround a carrier frequency wc and negligible elsewhere

)](cos[)()( ttwtatx c θ+=

Modulation converts a baseband signal to it’s passbandcounterpart, for ease of transmission (e.g. antenna size)

Demodulation extracts the baseband signal

Modulation/Demodulation Schemes = Modem

Important Aspects of ModemsImportant Aspects of Modems

Quality: SNR, or bit error rate

Bandwidth: Spectral Efficiency

Power Efficiency

전자정보대학 김영석 3-3

3.2 Analog Modulation: Amplitude Mod.For a baseband signal xb(t), the amplitude modulated signal is

Can be generated

ith i

twtxmAtx cBBcAM cos)](1[)( •+=

with a mixer

Passband signal has twice the bandwidth of the baseband signal

D d l ti : i ith i d LPF lDemodulation: mix with carrier and LPF, or use an envelope detector

전자정보대학 김영석 3-4

3.2.2 Phase and Frequency ModulationPhase and Frequency are related by dtdw /φ=

Phase modulated signal is expressed as (m=phase mod. Index)

F d l t d i l i d

)](cos[)( tmxtwAtx BBccPM +=

Frequency modulated signal is expressed as

])(cos[)( ∫ ∞−+=

tBBccFM dttxmtwAtx

전자정보대학 김영석 3-5

Modulation/DemodulationFrequency Modulation can be done by varying the freq. of an OSC with the baseband signal

Frequency Demodulation can be performed by freq. sensitive ckt, l b PLLcommonly by a PLL

전자정보대학 김영석 3-6

3.3 Digital ModulationMost modern wireless systems use digital modulation due to its advantages over analog modulation

Quality of RF systems is evaluated by BER(bit error rate). Modulation methods are compared according to their probability ofModulation methods are compared according to their probability of error (Pe)

전자정보대학 김영석 3-7

3.3.2 Basic conceptsBasis functions: any modulation scheme can be represented by a set of orthogonal basis functions

Def. orthogonal: 0)()(0

=∫sT

km dttt φφ

E.g. FSK(Frequency shift keying)

tttxFSK 2211 )()()( += φαφα

twttwtfunctionsbasisaforAandaforAwhere cc

FSK

2211

21

2211

cos)( ,cos)( 0 ]0 [ 1 ] 0[],[

)()()(

===

φφαα

φφ

Signal constellations: modulated waveforms can be viewed in terms of their coefficients and basis functions

Each basis functions contributes one axis to the plotEach basis functions contributes one axis to the plot

전자정보대학 김영석 3-8

3.3.2 Basic conceptsOptimum detection: It can be shown that the optimum detector is a correlator, which can be built by a multiplier (mixer) followed by an integrator (p(t)=pulse shape)

Integration produces an averaging that reduces the effects of the noise

전자정보대학 김영석 3-9

Digital ModulationBPSK(Binary Phase Shift Keying) wheretwAtx ccBPSK °=+= 180or 0)cos()( φφ

AcAcwheretwtx cBPSK

ccBPSK

−+== or cos)()()(αα

φφ

iG ihiddi ihfd ilhidbihih

)2

(0

,

NENE

QP bBPSKe =

전자정보대학 김영석 3-10

noiseGaussian whiteadditivetheofdensity spectralpower theisandbit per energy average theiswhere 0NEb

Digital ModulationBFSK(Binary Frequency Shift Keying) coscos)( 2211BPSK twtwtx += αα

]0 [ ] 0[] [ 21 cc AorAwhere =αα

iG ihiddi ihfd ilhidbihih

)(0

,

NENE

QP bBPSKe =

전자정보대학 김영석 3-11

noiseGaussian whiteadditivetheofdensity spectralpower theisandbit per energy average theiswhere 0NEb

Digital ModulationQuadrature Modulation: basis functions twtw cc sin and cos

(a) QPSK(Quadrature Phase Shift Keying)

Send 2 bits with one symbol][][][][][

sincos)( 21 ccQPSK

AAAAAAAAwhere

twtwtx +=

αα

αα

]-[-],-[,][- ],[] [ 21 cccccccc AAAAAAAAwhere =αα

noiseGaussian whiteadditive theofdensity spectralpower theis andbit per energy average theis where

)2

(

0

0,

NENE

QP

b

bBPSKe =

A problem with QPSK is the sharp transitions => OQPSK or π/4-QPSK

전자정보대학 김영석 3-12

5.2 Receiver ArchitecturesWhy not perform Channel Selection and Demodulation at the RF?

Tuneable Filters are harder to realize

Very High Q Channel Select Filter at fc is needed

• Q=fc/Δf• Ex: IS-54, 30kHz, 900MHz, 60dB Attenuation at 45kHz offset,

2nd-order LC filter Q=107 (Impossible)2nd-order LC filter Q=107 (Impossible)

Downconverting RF to IF before Channel Selection using a variable VCO => Use a fixed, low-Q filter (low fc, same Δf)

전자정보대학 김영석 3-13

5.2.1 Heterodyne Receivers

Herero=different dyne=mixHerero=different, dyne=mix

RF to IF(Intermediate Frequency) by Downconversion Mixing

High Q Filter is not required

Inclusion of LNA to lower the Noise Figure

전자정보대학 김영석 3-14

Problem of ImageSignal and Image are downconverted to the same frequency

How to suppress the Image: Image-Reject Filter

전자정보대학 김영석 3-15

IF Selection Trade-off between Image Rejection and Channel Selection

High IF: Better Image Rejection

Low IF: Better Channel Selection

Other factors: Availability, Physical Size of IF filtersOther factors Availability, Physical Size of IF filters

High or Low-side Injection

Low-side: Minimize LO freq

High-side: Reduced Tuning Range for the VCO

전자정보대학 김영석 3-16

Dual-IF TopologyTo Lessen the trade-off between Image Rejection and Channel Selection, a Dual-IF is used

1st-mixer converts to a relatively high IF to allow Image Rejection

2nd-mixer converts to a low IF to allow Channel Selection2nd-mixer converts to a low IF to allow Channel Selection

전자정보대학 김영석 3-17

5.2.2 Homodyne ReceiversHomodyne, Direct-Conversion, Zero-IF Architecture

Only LPF is required for channel selection

Topology (a) : Double-sideband AM (Same information on both sides of the spectrum)sides of the spectrum)

Topology (b) (Quatrature Downconversion): FM or PM (Different information on the different sides of the signal)

전자정보대학 김영석 3-18

Homodyne ReceiversAdvantages

No Image Filter

No IF SAW filter

Si l Chi R (IC)Single Chip Rx (IC)

Drawbacks

DC OffsetsDC Offsets

I/Q Mismatch

Even-order Distortion

Flicker Noise

전자정보대학 김영석 3-19

DC OffsetsDC Offset Voltages result from Self-mixing due to Poor isolation between LO and RF during downconversion

Example:

Input RF = 1μVrms (A) => Mix output RF = 30μVrms (30dB)Input RF = 1μVrms (A) => Mix output RF = 30μVrms (30dB)

LO = 0.63V (0dBm) => LO leakage = 0.63mV (60dB Isolation)

=> Mix output DC offset = 10mV (30dB gain)p g

=> Saturates the gain stages (50-70dB gain)

전자정보대학 김영석 3-20

Solutions for DC OffsetsHigh Pass Filtering to remove DC offsets

Coding is required to remove the information at DC

Low Cutoff Frequency => Large C

W t S tWastes Spectrum

Perform Periodic Offset Cancellation

Periodically sample the signal and cancel itPeriodically sample the signal and cancel it

Works best for TDMA Systems

Most common solution

전자정보대학 김영석 3-21

I-Q MismatchMismatch in gain and phase between I/Q LO signals => gain and phase error in the received signal constellation

Eg. for QPSK signal

1)i ()()( ±bhbRF

)2

cos()2

1(2)(:

1,)sin()cos()(:

,θε

++=

±=+=

twtvLO

QIbawheretwbtwatxRF

cILO

cc

)2

sin()2

1(2)(

)2

()2

()(

,

,

θεθε

θε−−= twtv cQLO

cILO

2cos)

21(

2sin)

21()(

2sin)

21(

2cos)

21()(:)(

,

,

θεθε

θεθε

−+−−=

+−+=

batx

batxmixingafterIF

QBB

IBB

Amplitude mismatch < 1dB, phase error < 5deg

전자정보대학 김영석 3-22

Even-Order DistortionLike 3rd-order distortion (IP3), 2nd-order Nonlinearity of Amp is problematic in homodyne

Two High-freq interferers generate a low-freq beat in the presence of even-order distortion => Direct feedthrough from RF to IF (low-of even order distortion > Direct feedthrough from RF to IF (lowfreq beat multiplied by coswLOt is translated to high freq: not a problem)

AA )()()(

twwAAtxtxtyoutputLNA

twAtwAtxRF

)cos(...)()()(:

)cos()cos()(:

212122

21

2211

−+=+=

+=

ααα

Differential circuit topologies will help suppress 2nd-order distortionDifferential circuit topologies will help suppress 2 order distortion

전자정보대학 김영석 3-23

Flicker NoiseDownconverted (after only amplified by LNA and mixer) low-freq signals are influenced by 1/f of next stages

Despite all of these difficulties homodyne receivers become moreDespite all of these difficulties, homodyne receivers become more common

Due to high levels of integration

Standard in some applications (eg. pagers)

전자정보대학 김영석 3-24

5.2.3 Image-Reject Receivers90deg phase shift = multiply the spectrum by G(w)=-j*sgn(w)

Eg. sinwt => -coswt (90deg phase shift)

jwtjwt ejejwt −+−=22

)sin(

jwtjwt eewt −−−=−21

21)cos(

22

전자정보대학 김영석 3-25

Hartley ArchitectureAssume low-side injection

)i ()i ()(

coscos)(

tA

tAt

wwwwtwAtwAtx

imRF

imLOLORF

imimRFRF

+

−=−+=

)sin(2

)sin(2

)sin(2

)sin(2

)(

twwA

twwA

twwtwwtx

imLOim

LORFRF

imLOim

RFLORF

A

−+−−=

−+−=

)cos(2

)cos(2

)(

)cos(2

)cos(2

)(

twwA

twwAtx

twwA

twwAtx

imLOim

RFLORF

B

imLOim

RFLORF

C

−+−=

−−−=

out) cancel components (Image)cos()()()(

22twwAtxtxtx RFLORFBCIF −=+=

전자정보대학 김영석 3-26

Hartley Architecture90 deg phase shift is replaced with +45 shift in one path and -45 shift in the other

전자정보대학 김영석 3-27

Hartley ArchitectureGain and phase mismatch will lead to incomplete image rejection

4)/(

)RatioRejection Image(

)cos()()( ,sin)( .22

21

θε

θε

+≈=

++==

LOim

LOLOLOLOLOLO

APP

IRR

twAtxtwAtxEg

In most RF applications, IRR=60-70dB is required

For typical matching in IC, gain mismatch=0.2-0.6dB, phase

4sigP

For typical matching in IC, gain mismatch .2 .6dB, phase imbalance=1-5deg, IRR=30-40dB

So, Image Reject Filter must still be used

전자정보대학 김영석 3-28

Weaver ArchitectureTo avoid 90 deg phase shift (problematic), an alternate image reject architecture can be used.

Secondary Image problem

전자정보대학 김영석 3-29

5.2.4 Digital-IF ReceiversAfter initial downconversion, use ADC to digitize the signal and perform second mixing and filtering in the digital domain

Advantages: No I-Q mismatch, maximum flexibility

전자정보대학 김영석 3-30

5.2.5 Subsampling ReceiversSubsample the RF signal at the Nyquist rate (low LO)

Filter the resulting output,

eliminating the need for a mixer

Merits: simplify the design of the LO

Demerits: Noise aliasing (Subsampling by a factor m => Increase the noise power by a factor 2m)

전자정보대학 김영석 3-31

5.3 Transmitter ArchitecturesTx architectures are generally much less varied, because noise, interference rejection, and band selectivity are more relaxed.

5.3.1 Direct-conversion Transmitter

LO pulling: PA output couples to LO, corrupting its output by pulling it to a different frequencypulling it to a different frequency

전자정보대학 김영석 3-32

5.3.2 Two-step TransmittersPA frequency (w1+w2) is different from LO frequency (w1) (no LO pulling)

전자정보대학 김영석 3-33

ReferencesBehzad Razavi, RF Microelectronics, Prentice Hall, Inc, 1998

전자정보대학 김영석 3-34