sip and future of semiconductor testsuch as flip-chip redistribution, redistribution interconnect,...
TRANSCRIPT
© 2014 Hanyang University ERICA, All rights reservedPage 1 June 25, 2014
SiP and Future of
Semiconductor Test(SiP와반도체테스트의미래)
Sung Chung, Research Professor
© 2014 Hanyang University ERICA, All rights reservedPage 2 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 3 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 4 June 25, 2014
Cloud Storage says NOOO!
© 2014 Hanyang University ERICA, All rights reservedPage 5 June 25, 2014
SiP Device says NOOOooop!
I am your father!
NOOOooop!
© 2014 Hanyang University ERICA, All rights reservedPage 6 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 7 June 25, 2014
System in Package (SiP) is a combination of
multiple active electronic components of
different functionality, assembled in a single
unit that provides multiple functions
associated with a system or sub-system.
A SiP may optionally contain passives,
MEMS, optical components and other
packages and devices.
Definition of SiP
“The next Step in Assembly and Packaging: System Level Integration in the package (SiP),” Assembly and Packaging Technical Working
Group for the ITRS, SiP White Paper V9.0
© 2014 Hanyang University ERICA, All rights reservedPage 8 June 25, 2014
3D-Packaging (3D-P):3D integration using “traditional” packaging technologies, such as wirebonding, package-on-package stacking or embedding in printed circuit boards.
3D-Wafer-Level-Packaging (3D-WLP):3D integration using wafer level packaging technologies, performed after wafer fabrication, such as flip-chip redistribution, redistribution interconnect, fan-in chip-size packaging, and fan-out reconstructed wafer chip-scale packaging
3D-Integrated-Circuit (3D-IC):3D approach using direct stacking of active devices. Interconnects are on the local on-chip interconnect levels. The 3D stack is characterized by a stack of front-end devices, combined with a common back-end interconnect stack.
3D-Stacked-Integrated-Circuit (3D-SIC):3D approach using direct interconnects between circuit blocks in different layers of the 3D die stack. Interconnects are on the global or intermediate on-chip interconnect levels. The 3D stack is characterized by a sequence of alternating front-end (devices) and back-end (interconnect) layers.
3D-System-on-chip (3D-SOC): SiPCircuit designed as a system-on-chip, SOC, but realized using multiple stacked die. 3D-interconnects directly connect circuit tiles in different die levels. These interconnects are at the level of global on-chip interconnects. This allows for extensive use/reuse of IP-blocks.
Definitions of Suggested Names
Source: ITRS 2011 Interconnect
© 2014 Hanyang University ERICA, All rights reservedPage 9 June 25, 2014
ITRS 3D IC TSV Roadmap 1
Global Level Interconnect
W2W, D2W, D2D Stacking2009 - 2012 2012 - 2015
Minimum TSV Diameter 4 - 8 µm 2 - 4 µm
Minimum TSV Pitch 8 - 16 µm 4 - 8 µm
Minimum TSV Height 20 - 50 µm 20 - 50 µm
Minimum TSV aspect ratio 5:1 - 10:1 µm 10:1 - 20:1 µm
Bonding Overlay Accuracy 1.0 - 1.5 µm 0.5 - 1.0 µm
Minimum Contact Pitch
Thermo-Compression10 µm 5 µm
Minimum contact pitch
Solder Micro-Bump20 µm 10 µm
© 2014 Hanyang University ERICA, All rights reservedPage 10 June 25, 2014
ITRS 3D IC TSV Roadmap 2
Intermediate Level Interconnect
W2W Stacking2009 - 2012 2012 - 2015
Minimum TSV Diameter 1 - 2 µm 0.8 - 1.5 µm
Minimum TSV Pitch 2 - 4 µm 1.6 - 3.0 µm
Minimum TSV Height 6 - 10 µm 6 - 10 µm
Minimum TSV Aspect Ratio 5:1 - 10:1 µm 10:1 - 20:1 µm
Bonding Overlay Accuracy 1.0 - 1.5 µm 0.5 - 1.0 µm
Minimum Contact Pitch 2 - 3 µm 2 - 3 µm
© 2014 Hanyang University ERICA, All rights reservedPage 11 June 25, 2014
2.5D/3D SoC vs. SiP Packaging Roadmap
“3DIC & TSV interconnects 2012 Business update,” Yole Development, Semicon Taiwan 2012
© 2014 Hanyang University ERICA, All rights reservedPage 12 June 25, 2014
Comparison of SoC/SiP Architecture 1
Market and Financial Issues
Item SiP SoC
Relative NRE cost 1X 4-10X
Time to Market 3 –6 months 6 –24 months
Relative Unit Cost 1X 0.2 – 0.8X
“The next Step in Assembly and Packaging: System Level Integration in the package (SiP),” Assembly and Packaging Technical Working Group for the
ITRS, SiP White Paper V9.0
© 2014 Hanyang University ERICA, All rights reservedPage 13 June 25, 2014
Comparison of SoC/SiP Architecture 2
Technical FeaturesSiP SoC
Pros
Different front end technologiesGaAs, InP, Si, SiGe, etc.
Better yields at maturity this depends upon complexity)
Different device generations Greater miniaturization
Re-use of common devices Improved performance
Reduced size vs. conventional packaging Lower cost in volume
Active & passive devices can be embedded CAD systems automate interconnect design
Individual components can be upgraded Higher interconnect density
Better yields for smaller chip sets Higher reliability (not true for very large die)
Individual chips can be redesigned cheaper Simple logistics
Noise & crosstalk can be isolated better
Faster time to market
“The next Step in Assembly and Packaging: System Level Integration in the package (SiP),” Assembly and Packaging Technical Working Group for the
ITRS, SiP White Paper V9.0
© 2014 Hanyang University ERICA, All rights reservedPage 14 June 25, 2014
Comparison of SoC/SiP Architecture 3
Technical Features
SiP SoC
Cons
More complex assembly Difficult to change
More complex procurement &
logisticsSingle source
Power density for stacked die may
be too high
Product capabilities limited by chip
technology selected
Design Tools may not be adequateYields limited in very complex, large
chips
High NRE cost
“The next Step in Assembly and Packaging: System Level Integration in the package (SiP),” Assembly and Packaging Technical Working Group for the
ITRS, SiP White Paper V9.0
© 2014 Hanyang University ERICA, All rights reservedPage 15 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 16 June 25, 2014
IMEC 3D TSV Scaling Roadmap
Eric Beyne, Herve Ribot, Jurgen Wolf, Francoise von Trapp, “3D TSV Without Limits Webinar,” 3D InCites, 2013
© 2014 Hanyang University ERICA, All rights reservedPage 17 June 25, 2014
Resistance-capacitance (RC) of a die-to-die (d2d) via related to a full via stack and signal propagation delays for a d2d via, 1 mm of wire, and a fan-out-of-four (FO4) delay
RC Characteristics of TSV
Loh, Gabriel H. et al, “Processor Design in 3D Die-Stacking Technologies,” in IEEE Micro, vol.27, issue 3, pp. 31-48, 2007
© 2014 Hanyang University ERICA, All rights reservedPage 18 June 25, 2014
For 22nm technology, delay will be 100ps for the 300µm 2D wire length. In contrast, the TSV with 22nm inverter has about 7ps delay.
TSV Delay Comparison
Ryusuke Egawa et al, “Poster: Exploring Design Space of a 3D Stacked Vector Cache - Designing a 3D Stacked Vector
Cache using Conventional EDA Tools,” IEEE SCC 2012, DOI=10.1109/SC.Companion.2012.271
© 2014 Hanyang University ERICA, All rights reservedPage 19 June 25, 2014
Small RC Delay1mm top metal
delay: 225 ps
One TSV: 8 ps
Through 20 layers:
12 ps
High density
Interconnect:1 Million TSVs/mm2
Delay: TSV vs. 2D wiring
Hiroaki Kobayashi, “A Multi-Vectorcore Architecture with 3D Die-Stacking Technology,” 14th
Teraflop Workshop, Stuttgart, Germany Dec. 2011
© 2014 Hanyang University ERICA, All rights reservedPage 20 June 25, 2014
Power: TSV vs. 2D wiring
Hiroaki Kobayashi, “A Multi-Vectorcore Architecture with 3D Die-Stacking Technology,” 14th Teraflop Workshop, Stuttgart, Germany Dec. 2011
© 2014 Hanyang University ERICA, All rights reservedPage 21 June 25, 2014
For 45nm technology, equivalent 2D wire length to TSV is 100
µm for the TSV with 50µm long and 8µm in diameter
Long wire length and top metal wire increases wire delay
substantially
TSV and 2D Wire Delay
Ryusuke Egawa et al, “Poster: Exploring Design Space of a 3D Stacked Vector Cache - Designing a 3D Stacked Vector Cache using Conventional EDA
Tools,” IEEE SCC 2012, DOI=10.1109/SC.Companion.2012.271
Diameter
(µm)
Length
(µm)
Resistance
(mΩ)
Capacitance
(fF)
8 50 17 94
4 50 68 50
2 50 274 27
1 50 1095 16
0.5 50 4380 10
© 2014 Hanyang University ERICA, All rights reservedPage 22 June 25, 2014
Within the area of 5 µm diameter TSV, more than
100s of gate in 45 nm technology can be placed
Number of Gates in one TSV Area
Ryusuke Egawa et al, “Poster: Exploring Design Space of a 3D Stacked Vector Cache - Designing a 3D
Stacked Vector Cache using Conventional EDA Tools,” IEEE SCC 2012, DOI=10.1109/SC.Companion.2012.271
High density
interconnect:1 Million TSVs/mm2
© 2014 Hanyang University ERICA, All rights reservedPage 23 June 25, 2014
TSV Structure and Fault Model
Metal
TSV
Metal
SiO2 TSV Insulator
Silicon Substrate
Leakage Current LiTSV
RTSV
CTSV
Cu
Adopted from Yi Lou & Zhuo Yan, et al.,“Comparing Through-Silicon-Via (TSV)
Void/Pinhole Defect Self-Test Methods,” J of Electron Test (2012) 28:27–38
Example: 32 nm IMEC TSV
TSV Diameter: 5 µm
Pitch min.: 10 µm
Wafer: 25 µm thick
TSV Resistance: 20 mΩ
CTSV: 37 fF at 1 MHzLeakage LiTSV: 1 pA
Leakage RTSV: > 850 MΩ
Density: 105-8 cm2
Fault ModelsBreak defect
Pin hole defect
Open (Void) defect
CMOS
© 2014 Hanyang University ERICA, All rights reservedPage 24 June 25, 2014
SiO2 TSV Insulator
TSV Fault Models
Metal 1 Metal 1
SiliconSubstrate
TSVTSV
Full open defect
Break defect
Pin-hole Defect
Void Defect
Pin hole defect in SiO2 insulationCreate resistive short
Leakage current increases
Void defect (Open defect) in Cu:Signal is disconnected or
form a resistive path
Create data path signal integrity issues
Partially open defects are
difficult to detect
Break defect in TSV CuCapacitance is reduced and
form a resistive path
May create resistive short
Partially open break defect causes
electro migration and causes latent TSV
failure
Create data path signal integrity issues
Cu Cu
© 2014 Hanyang University ERICA, All rights reservedPage 25 June 25, 2014
TSV
TSV
TSV
TSV
TSV
TSV
Spare TSV and Repair Types
Shared TSV Method
Signal Shifting Method
Dedicated Spare TSV Method
Signal Switching Method
Signal Routing Method
Combination Shared TSV Method
Basic TSV Redundancy Mechanism
TSV
TSV
TSV
TSV
TSV
2 to 1 Mux
3 to 1 Mux
3 to 1 Mux
2 to 1 Mux
3 to 1 Mux
3 to 1 Mux
3 to 1 Mux
3 to 1 Mux
TSV
TSV
TSV
TSV
TSV
TSV
1 to 4Demux
1 to 4Demux
1 to 4Demux
1 to 4Demux
4 to 1 Mux
4 to 1 Mux
4 to 1 Mux
4 to 1 Mux
TSV
TSV
TSV
TSV
TSV
TSV
TSV
TSV
TSV
TSV
TSV
TSV
1 2 3 4 5 6 7 8 9S1 S2 S3
TSV
S0
© 2014 Hanyang University ERICA, All rights reservedPage 26 June 25, 2014
Quadruple TSV
16:4 TSV
Redundant TSVs & Stacking Yield
H. Hashimoto et al.,, “Highly Efficient TSV Repair Technology for Resilient 3-D Stacked Multicore Processor
System,” IEEE 3DIC Conference, 2013. DOI 10.1109/3DIC.2013.6702338
Stacking yield is a function of given TSV repair strategy
© 2014 Hanyang University ERICA, All rights reservedPage 27 June 25, 2014
Mitsumasa Koyanagi et al., “Three-Dimensional VLSI System with Self-Restoration Function,” JST
International Symposium on Dependable VLSI Systems 2013
Overhead and Signal Span
© 2014 Hanyang University ERICA, All rights reservedPage 28 June 25, 2014
16 signals and 4 Spares
Comparison of TSV Repair Yield
Mitsumasa Koyanagi, “Architecture and Circuits for Dependable 3D-
VLSI,” Tohoku University Sendai, Japan
© 2014 Hanyang University ERICA, All rights reservedPage 29 June 25, 2014
TSV Characteristics
Etienne Sicard et al., “Introducing 32 nm technology in Microwind35,”
Microwind Application Note
Small RC time delayOne TSV: 8 ps12 TSV layers: 12 ps32nm gate delay: 5 to 6 ps
1mm top metal delay: 225 ps
Electrical parameterTSV Capacitance: 37 fF at 1 MHzLeakage Resistance: > 850 MΩLeakage Current: 1 pA
Testing single pre-bond TSV
for all parametric value
becomes very difficult
© 2014 Hanyang University ERICA, All rights reservedPage 30 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 31 June 25, 2014
JTAG Standard covers all
different variations of current
/ future device needs
JTAG Standard will be fully mature
when 1838 is completed
1149.1 remains backbone of all
variations of JTAG Standard
May expect subsequent update of
existing standard once 1687 and
1838 become active
Genealogy of 1149.1 JTAG Family
AnalogDigital
1149.1 1149.3
JTAG – 1149.1
1149.2 1149.4
1149.5
1149.6
1532
1687
1838
1149.1
System
Device/PCB
AC Coupled
Compact Pin
ISP
3D IC
Memory
1500
Core Test
Instrument
1581Toggle
1149.8.1
1149.7
European JTAG
Active Standard
In Development
Withdrawn/Inactive
DC Coupled
Standard VariationDirectly or Indirectly Supporting Standard
1999 2010
2000 2002
2003 2003
2005 2005
2009 2009
2011 2011
2012 2012
2014? 2014?
1990 1993
1149.11990 2001
2001 2013
1149.1
First Lease
Active Version
EJTAG
© 2014 Hanyang University ERICA, All rights reservedPage 32 June 25, 2014
P1687 iJTAG (Instrument): in vallot process..Standard methodology for access to embedded test and debug features through IEEE Std. 1149.1This Standard provides a method of adding additional lines and functionality to the JTAG TAP to enable far greater levels of internal testing to be achieved using internal instrumentation
iJTAGIEEE P1687
1. Integrated into existing ICT2. ICT system > P1687 test
solution3. P1687 test solution >
functional test4. Integrated into existing
functional test
Key Differences between JTAG and iJTAG
JTAG iJTAG
Control of internal IPAd hoc method, vendor specific
Standard protocol
External interface to internal instrument and 3rd-party IP
Need information from instrument vendor
Plug-and-play and vendor independent
Coolness Old and boring Fancy and New
Instrument access through hierarchical logic structure
Must be manuallydefined of the JTAG interface
Automated retargeting from TAP to instrument through logic hierarchy
Register size Fixed per instruction Flexible
© 2014 Hanyang University ERICA, All rights reservedPage 33 June 25, 2014
Analysis of IEEE P1687 network
Farrokh Ghani Zadegan, et al., “Test Time Analysis for IEEE P1687”, IEEE 19th ATS 2010, Shanghai, China, Dec. 2010
The control data for the Segment Insertion Bit (SIB)
is transported on the same wire as test data
Control data is transferred to the status register
when the JTAG state machine does ”apply and
capture”
“CUC”: Apply Capture Update Cycle: 5 clock
cycles in the FSM
Key is to reduce Test Application Time (TAT)
© 2014 Hanyang University ERICA, All rights reservedPage 34 June 25, 2014
3D TestIEEE P1838
P1838 3D Device Test & DFTStandard for Test Access Architecture
for Three-Dimensional Stacked Integrated Circuits
Scope: The proposed standard is a ‘die-centric’ standard; it applies to a
die that is pre-destined to be part of a multi-die stack and such a die
can be compliant (or not compliant) to the standard
Two Standardized Components3D Test Wrapper hardware per die
Description + description language
Leverage existing DFT wherever Applicable/AppropriateTest access ports: utilize IEEE Std. 1149.x
On-die design-for-test: utilize IEEE Std. 1500
On-die design-for-debug: Utilize IEEE P1687
Web Page http://grouper.ieee.org/groups/3Dtest/
© 2014 Hanyang University ERICA, All rights reservedPage 35 June 25, 2014
IEEE1149.7
P1838 uses the largest number of standards to solve 3D test challenges
Relationship between All Standards
IEEE1149.11149.6
IEEE1581
Nexus5001
IEEE1532
EJTAG
SJTAG
IEEEP1687
IEEE1500
IEEEP1838
© 2014 Hanyang University ERICA, All rights reservedPage 36 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 37 June 25, 2014
What to Test? Determine appropriate test contentPre-bond (known good die, pretty good die): Types of tests
for TSVs and die logic
Post-bond (known good stack): TSVs, die logic (cores and
memories), tests for failing new defect types
How to Test? Determine test delivery pathwaysUse of BIST, BOST, or probe access based test
Scan chains, wrappers, test access mechanisms, test
scheduling, debug, diagnosis, and repair
Clock, test data, and TSV repair and configuration
When to Test? Select test flows: cost and benefit
trade-off analysis
SiP Typical Test Objectives
© 2014 Hanyang University ERICA, All rights reservedPage 38 June 25, 2014
3D TestingTesting at individual die level
Determine Known Good Die (KGD) for stackingTesting at 3D stack level
Final stack test
Individual die test optionsNo TestBuilt-In-Self-Test (BIST) and Built-Out-Self-Test (BOST)Reduced-Pad-Count-Test (RPCT)
3D Testing assumptionsFinal, post-stacking test performed in all casesTest Bottom die
Avoid stacking cost if die is faultyScan chains in place on all stacked dice
To be utilized in the final, post-stacking test
Solving new challenges as we goNo available Test Standard No mature EDA tool setWe haven’t uncovered all defect types yetTSV and interposer pre and mid-bond test pose most challenges
Test is Considered “3D Challenge No. 1”
© 2014 Hanyang University ERICA, All rights reservedPage 39 June 25, 2014
How Long is Nano-Second?
Maximum Limiting Distance of
electricity can travel in nano
second (10-9 second) is
29.97 cm.
In micro second, electricity can
travel 299.79 meter.
Grace Murray Hopper: Dec. 9, 1906 to Jan. 1, 1992
© 2014 Hanyang University ERICA, All rights reservedPage 40 June 25, 2014
Light Speed and High Speed
During one Pico Second, light can travel the length of 5.9 TSVsWhen light passes through 2.99 mm, electrons already travel through one TSV
(Based on 5µm diameter and 50µm long TSV with 8 ps signal delay)
Time Variables Speed of Light (Metric) Frequency
One Second 299,792,485 m/sec Hz
Micro-Second (10-6) 299.792 m/micro-sec MHz (106)
Nano-Second (10-9) 29.97 cm/nano-sec GHz (109)
100 Pico-Second (10-10) 2.99 cm/100 pico-sec 10 GHz (1010)
10 Pico-Second (10-11) 2.99 mm /10 pico-sec 100 GHz (1011)
Pico-Second (10-12)0.29 mm/pico-sec
299.79 µm/pico-secTHz (1012)
© 2014 Hanyang University ERICA, All rights reservedPage 41 June 25, 2014
Scaling analysis of the RLCG of the TSV pair according to ITRS
TSV RLCG Scaling with ITRS
C. Xu, H. Li, R. Suaya, and K. Banerjee, “Compact AC Modeling and Performance Analysis of Through-Silicon
Vias in 3-D ICs,” IEEE Trans. Electron Devices, vol. 57, no. 12, December 2010.
CNT: Carbon Nano Tube
SWCNT: Single Well CNT
MWCNT: Multi Well CNT
© 2014 Hanyang University ERICA, All rights reservedPage 42 June 25, 2014
SiP Test Needs Better Solution
TSV and Interposer impose new challenges for SiP test
Defect affects differently during pre-bond & post-bond testDuring pre-bond test
Defect makes R and C smallerDefect makes RC time constant shorter
During post-bond testDefect increase electrical maskingDefect makes RC time constant longer
Test parameter changes as stacking continuesIncrease delay and electrical maskingTSV interference and cross talk
Test structure changes as stacking continuesDifferent test for the same level of stack
TSV involves repairRepair changes timing and test resourcesTake PGD instead of KGD
© 2014 Hanyang University ERICA, All rights reservedPage 43 June 25, 2014
SiP Test Needs Better Solution: cont.
Ownership of repair and cost are not fully definedVague ownership of process improvement and requirement
Complex logistics
Stack has process dependencyParametric values may vary per each stack
Accuracy of global BIST (BOST) may have PVT dependency
Prefer to generate test clock at each stack to limit variation of test
accuracy
Standard doesn’t solve TSV test itself, but…… it’ll standardize test access and execution. We need to develop
test of our own.
It’s at least 3 year out to have fully approved IEEE1838 3D Test
Remember, TSV & Interposer opened door to SiP
© 2014 Hanyang University ERICA, All rights reservedPage 44 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 45 June 25, 2014
TSV / Interposer Delay Test
Pre-bond and post-bond need
different test methodRLCG* values of pre-bond and post-bond are different
Resistance RTSV,Inductance LTSV,Capacitance CTSV,Conductance GTSV
Parametric values of pre-bond and post-bond are different
Pre-bond defect decrease RC time constantPost-bond defect increase RC timeconstant
Time and frequency
measurement techniques are the
most common TSV delay test
FabDefect,
Fault
• Ideal TSV (Defect Free)
• Inductance LTSV, Capacitance CTSV, Resistance RTSV, Conductance GTSV
Test Activation
• Apply Stimulus (Defect → Error)
• Test pattern: [ i, v] in t or f domain
• Target fault: structural, functional
Compare Failure
• Check Response (Error → Failure)
• Logic check: go-no-go, logic state
• Parametric value: [ i, v] in t or f domain
Pass,
Fail
• Pass-Fail (Repair Failure)
• Repair with redundancy and retest
• Unrepairable defect yield loss
* C. Xu, H. Li, R. Suaya, and K. Banerjee, “Compact AC Modeling and
Performance Analysis of Through-Silicon Vias in 3-D ICs,” IEEE Trans.
Electron Devices, vol. 57, no. 12, December 2010.
Stacking alters AC/DC test parameters
© 2014 Hanyang University ERICA, All rights reservedPage 46 June 25, 2014
Standard CTSV: 37 fFResolution: 3.3 fF (8.9%)
Response speed: 0.045 ns/fF
Standard RTSV: 850 MΩ
Leakage LiTSV: 1 pA
Test Method:
Sense Amplification
Target Fault: Void Defect
(capacitance dominant defect)
Measure Capacitance with Sense Amplification
Discharge time is measured by sense amplifier
(inverter threshold voltage)
Use large width PMOS (charge time) and long
channel NMOS (resistance)
Y. Lou, Z. Yan, F. Zhang, and P.D. Franzon, "Comparing Through- Silicon-Via (TSV) Void/Pinhole Defect Self-Test Methods", presented at J. Electronic Testing, 2012, pp.27-38.
L2VCC (leakage to voltage conversion and then comparison)
© 2014 Hanyang University ERICA, All rights reservedPage 47 June 25, 2014
Modified Sense Amplification
Standard CTSV: 37 fF
Standard RTSV: 850 MΩDetectable Equivalent Leakage
Resistance: 2 kΩ
Leakage LiTSV: 1 pA
Target Fault: Pin Hole Defect(resistance dominant defect)
Fail or long to pre-charge TSV from leakage
Measure equivalent leakage resistance
Discharge delay is measured by sense amplifier (inverter threshold voltage)
Y. Lou, Z. Yan, F. Zhang, and P.D. Franzon, "Comparing Through- Silicon-Via (TSV) Void/Pinhole Defect Self-Test Methods", presented at J. Electronic Testing, 2012, pp.27-38.
L2VCC (leakage to voltage conversion and then comparison)
© 2014 Hanyang University ERICA, All rights reservedPage 48 June 25, 2014
Standard CTSV : 37 fF
Standard RTSV : 850 MΩ
Resolution: 10 MΩ (1.1%)
Response speed: 0.4 V/MΩ
Leakage LiTSV : 1 pA
Test Method:
Leakage Current Sensor
Target Fault: Leakage Pin-Hole, Sidewall Short Defect
(resistance dominant defect)
Controlled (long, 50µS to charge 37 fF) test initialization process
Voltage is measures against threshold voltage of comparator
Consider gate leakage for accuracy
TSV R=20mΩ, LiTSV = 1pA, Inverter Rinput = 3GΩ, Inverter LiD-S = 1nA,
PMOS Li = 1nA, RTSV = 850 ΩM
Y. Lou, Z. Yan, F. Zhang, and P.D. Franzon, "Comparing Through- Silicon-Via (TSV) Void/Pinhole Defect Self-Test Methods", presented at J. Electronic Testing, 2012, pp.27-38.
L2VCC (leakage to voltage conversion and then comparison)
© 2014 Hanyang University ERICA, All rights reservedPage 49 June 25, 2014
Standard CTSV: 37 fF
Resolution: 0.16 fF (0.4%)
Response speed: 250 ns/fF
Standard RTSV: 850 MΩ
Leakage LiTSV: 1 pA
Test Method:
Capacitance Bridge
Target Fault: Full Open Void Defect(capacitance dominant defect)TSV capacitance is compared with an on-chip referenceTSV capacitance is proportional to delay time from the reset signal goes low to output risesRequire ring OSC (2 GHz) generating pseudo sinusoidal waveform with low-pass RC filter (4kΩ, 50fP)
Y. Lou, Z. Yan, F. Zhang, and P.D. Franzon, "Comparing Through- Silicon-Via (TSV) Void/Pinhole Defect Self-Test Methods", presented at J. Electronic Testing, 2012, pp.27-38.
© 2014 Hanyang University ERICA, All rights reservedPage 50 June 25, 2014
BIST for Pre-bond TSV
Zimouche, H. “A BIST method for TSVs pre-bond test,” Design and Test Symposium (IDT), 8th International, 2013
Full-open break faultTSV discharge 7.5 ns
Delay 1: 7.5 ns
Delay 2: 6.5 ns
Pin-hole defectDefect < 10MΩ
Highly PVT dependent
Pre-bond test only
L2VCC (leakage to voltage conversion
and then comparison)
© 2014 Hanyang University ERICA, All rights reservedPage 51 June 25, 2014
L2VCC 2 (leakage to voltage conversion and then comparison)
CAF-WAS 1 (charge-up-and-float, wait-and-sample)
CAF-WAS Wait time = (CTSV × 0.5VDD) / LTT
TSV pre-bond test
Require programmable delay lineLeakage of 0.125 µA (160 ns) to 16 µA (1.25 ns)1024 to 8 buffer delays
Delay based Leakage Test
Shi-Yu Huang et al., "Programmable Leakage Test and Binning for TSVs With Self-Timed Timing Control", Computer-Aided Design of Integrated Circuits and Systems, IEEE Trans. on, On page(s): 1265 - 1273 Volume: 32, Issue: 8, Aug. 2013
© 2014 Hanyang University ERICA, All rights reservedPage 52 June 25, 2014
CAF-WAS “Charge-up-and-float, Wait-and-
sample” method
Wait time: CTSV × 0.5VDD) / LTTLTT: Leakage Test ThresholdLTT is 1 µA to 10 µA for the standard I/O pinFor 1 µA LTT is as in the standard IO pin leakage test, the wait-time will be about 20 ns (40fF x 0.5V)/1µA ).
TSV pre-bond testDetectable capacitance range: 40fF~242fF
Detectable leakage range: 1 µA to 128 µA
Useful to perform leakage binning then test
Require programmable delay line for wait-time
generationLeakage of 1 µA (20 ns) to 128 µA (0.156 ns)64 to a buffer delaysEach ‘INVX2’ cell has a delay of 15.55 ps (unit delay)
Programmable Leakage Test
Y.-H. Lin, S.-Y. Huang,K.-H. Tsai, and W.-T. Cheng, "Programmable Leakage Test and Binning for TSVs", Proc. of IEEE Asian Test Symp. (ATS), pp. 43-48, Nov. 2012.
© 2014 Hanyang University ERICA, All rights reservedPage 53 June 25, 2014
Test leakage fault with ring oscillatorUse variable output threshold analysis methodDelays are measured by on chip counterHigh overhead: for 32768 TSV has normalized percentage (TSV) of 39.9%Highly accurate (±1%)Impact of process variation is 7.9%
99.9% confidence level that TSV delay is within the maximum error bound
Parametric Delay Test
Lin, Y.-H. et al., "Parametric Delay Test of Post-Bond Through-Silicon Vias in 3-D ICs via Variable Output Thresholding Analysis", IEEE Trans. on Computer-Aided Design of Integrated Circuits and Systems, pp 737 - 747 Vol. 32, Iss. 5, May 2013
ΔTST = TST – TREF
RTSV=60 mΩ, LTSV=2.8 pH, CTSV=40 fF, GTSV=0.1 fΩ-1.
TST: Schmitt-Trigger delay, TREF: Normal delay
© 2014 Hanyang University ERICA, All rights reservedPage 54 June 25, 2014
Interposer Wire Delay Test
Interposer wire delay test using
boundary scanLaunch short pulse from the driver
Electrical Making Effect:Masking affects pulse height at the
receiver due to RC time constant
Pulse height (driver output current)
and pulse width must be controlled
Test Accuracy Dependence: Pulse width must be pre-determined
Receiver threshold (temp
dependent) affect test accuracy
Wire length (RC time constant)
affects test accuracy
Shi-Yu Huang et al., “At-speed BIST for interposer wires supporting on-the-spot diagnosis,” IEEE 19th IOLTS, 2013
© 2014 Hanyang University ERICA, All rights reservedPage 55 June 25, 2014
Interposer Wire Delay Test: cont.
Shi-Yu Huang et al., “At-speed BIST for interposer wires supporting on-the-spot diagnosis,” IEEE 19th IOLTS, 2013
© 2014 Hanyang University ERICA, All rights reservedPage 56 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples: Boeing 787
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 57 June 25, 2014
About Boeing 787
A midsize, wide-body, twin-
engine jet airliner
The most fuel-efficient airliner
ever built
First to use composite materials as
a primary material
First Unveiled July 2007 (677 on
order)
Maiden Flight Dec. 2009
Entered commercial service
2011
Originally planned to enter
service in 2008
Boeing 787 requires about 6.5 million
lines of code to operate its avionics
and onboard support systems.
© 2014 Hanyang University ERICA, All rights reservedPage 58 June 25, 2014
B787: Million-Lines-Of-Code
http://www.informationisbeautiful.net/visualizations/million-lines-of-code/
© 2014 Hanyang University ERICA, All rights reservedPage 59 June 25, 2014
July 12 2013: Ethiopian Airlines plane catches fire on the runway at Heathrow, forcing the closure of the whole airportJanuary 15 2013: A flight made an emergency landing in Japan after a smoke alarm went off. The string of incidents led to regulators ordering a global grounding of the entire Dreamliner fleet, which lasted for four monthsJanuary 7 2013: A Dreamliner catches fire after dropping off 183 passengers and crew in Boston when a battery in the jet's auxiliary power system overheated.Jan 8, 2013: fuel leak cancels international flightJan 9, 2013: wiring problems reported
Nightmares: Highlighted Incident
http://www.dailymail.co.uk/news/article-2260733/Aviation-regulators-demand-safety-review-Boeing-787-Dreamliners-series-mechanical-structural-failures
© 2014 Hanyang University ERICA, All rights reservedPage 60 June 25, 2014
Erratic behavior of APU and
eventual failure of APU
Exhaustive DFTA (Dynamic Fault
Tree Analysis) couldn’t prevent
these failures
Controller and battery supplier
are from different country
Problem stems from widespread
use of outsourcing
Auxiliary Power Unit (APU) Failure
Nicholas Williard et al, “Lessons Learned from the 787 Dreamliner Issue on Lithium-Ion Battery Reliability,” doi:10.3390/en6094682
APU exhibits internal short on one of
battery cells
© 2014 Hanyang University ERICA, All rights reservedPage 61 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples: Toyota Camry
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 62 June 25, 2014
Dual Core
Lock-Step
Sphere of Replication
ECC Protected
Memories
Replicated Peripherals
Fault Collection Control
Unit (FCCU)
JTAG and Nexus
BIST
Automotive MCU Overview
Markus Baumeister, “Addressing Safety Standard Requirements for IEC61508 (SIL3) and ISO26262 (ASIL-D) with the
MPC5643L 32-bit Power Architecture Microcontroller,” Freescale
© 2014 Hanyang University ERICA, All rights reservedPage 63 June 25, 2014
It’s everywhereSmart Devices: “The Internet of Things”Becoming necessity for everyday life for everyone
It’s in your car and very complex
SW development is about 13 -15% of total development cost Estimates that 50% of car warranty costs are related to electronics and their embedded SW
It costs automakers around $350 per car in the US
It has security risksHighly connected to the internetIncreasing security and safety incidence
Can hijack your car from the smart phone
Embedded System
An average modern high-end car runs about 100
MLOC on 70 to 100 Micro Processor based ECUs
© 2014 Hanyang University ERICA, All rights reservedPage 64 June 25, 2014
Automotive: Million-Lines-Of-Code
http://www.informationisbeautiful.net/visualizations/million-lines-of-code/
© 2014 Hanyang University ERICA, All rights reservedPage 65 June 25, 2014
2006: Oklahoma, Tragic crash of a Toyota Camry which supposedly exhibited Unintended Acceleration (UA)Several other UA followed, resulting in several lawsuits
NASA’s Engineering and Safety Center investigated the incidentFound two possible failures of the ETCS that could cause the car to exhibit UA without the onboard diagnostics system detecting itNo direct evidence that those failures caused the accidents
Embedded systems experts reviewed the source codesToyota’s source code is defective, contains bugs, including bugs that can cause UAEven just the death of Task X by itself can cause loss of throttle control –throttle continues to power the engine at high RPM
The fail safes Toyota did install have gaps, are inadequate to detect all cases of UA can occur via software
Toyota’s fail safe measures are defective and inadequate
Dec. 2012: More than $1 Billion settlement by Toyota
The Toyota Camry UA Incident
© 2014 Hanyang University ERICA, All rights reservedPage 66 June 25, 2014
Detail: Task Death
The RTOS' critical internal data structures
The final result of all this firmware- the TargetThrottleAngleglobal variable.
Michael Barr “2005 Camry L4 Software Analysis,” BARR Group
© 2014 Hanyang University ERICA, All rights reservedPage 67 June 25, 2014
Representative of task death
in real-world
Dead task also monitors
accelerator pedal, so loss of
throttle controlConfirmed in tests
When this task’s death
begins with brake press (any
amount), driver must fully
remove foot from brake to end UA
Confirmed in tests
Detail: Test Confirmation
Michael Barr “2005 Camry L4 Software Analysis,” BARR Group
Single bit flip in the RTOS’s (unprotected) critical
data structures could kill tasks
© 2014 Hanyang University ERICA, All rights reservedPage 68 June 25, 2014
Soft-error & Automobiles
Mar,2010 - NHTSA enlisted
NASA Engineering and
Safety Center (NESC) to
investigate “Unintended
Acceleration”
Apr,2011 – NESC
discounts SEU in its report to
NHTSA stating that the ICs
manufactured using SOI
(Silicon-on-insulator)
technology
As per AEC-Q100 standard,
SEU testing required for
automobile electronics with
RAM > 1Mb
© 2014 Hanyang University ERICA, All rights reservedPage 69 June 25, 2014
Opening Remark
Definition and overview of SiP Technology
Characteristics of TSV in SiP
JTAG and SiP Test Standards
SiP Test Challenges: TSV
TSV and Interposer Test Examples
Recent System Failure Examples
Conclusions
Q and A
Presentation Outline
© 2014 Hanyang University ERICA, All rights reservedPage 70 June 25, 2014
Presented overview of SiP test challenges, especially
from TSV and interposer test view point
Reviewed difficulties related to TSV and interposer test
during pre, mid, and post-bond phases
Reviewed that the stacking modifies not only test data
and test architectures, but also changes test parameter
for pre, mid, and post-bond phases
Understood our challenges to manage SiP test are
mostly from TSV and related interposer necessary for the
stacking
Reviewed of existing legacy MCU and complex system
failures to understand challenged associated with
logistics and outsourcing
Conclusions
© 2014 Hanyang University ERICA, All rights reservedPage 71 June 25, 2014
“…. But there are also unknown unknowns — the ones
we don’t know we don’t know”
Donald Rumsfeld
Remember the unknown unknowns
Thank you!