Technology Development Methodology – CMOS as a game-changer
1.
Presenta)on
at
the
6th
Workshop
of
the
“IIPCC
HK
Chapter
IP
Workshop
Series:
IP’s
Role
in
Entrepreneurship
&
Innova)on”
(in
HKSTP
on
February
11,
2015)
Prepared
by:
Al
Kwok
(郭灿辉)
Governor
&
Co-‐founder,
IIPCC
Hong
Kong
Chapter
Governor
&
Co-‐founder,
HKIURCA
(⾹香港產學研合作促進會)
Member,
HKSAR
“IP
Trading”
Working
Group
Governor
&
Co-‐founder,
Savantas
Policy
Ins)tute
President,
CASPA
(华美半导体协会)
PRD
Chapter
Former
VP
&
CIPO,
NetLogic
Microsystems
(“NETL”)
Technology Development Methodology - CMOS 1
Technology Development
Methodology – Making CMOS
the Game-Changer!
Feb. 11, 2015
2. Acknowledgement
n Innovations in IC industry have been
collaborative efforts.
n I am indebted to my colleagues and co-
workers at IDT for their assistances and
supports for developing my early career to
acquire the needed domain expertise as a
“Yield Guru” to be a pioneer launching the
Information Age – driven by the Moore’s
Law and Metcalfe’s Law scalability.
Feb. 11, 2015 Technology Development Methodology - CMOS 2
3. Outline
n The Impact of CMOS: How & Why?
n As basic to “Information Age” as “Wheel” to “Early Civilization”
n IDT (founded by Indian & Chinese, first in Silicon Valley) made history
n The Basic in Technology Development
n The Big Picture: How it fits in? From Corporate and Individual perspectives
n Process, Building Blocks, Organization/Lay-out
n Supporting Equipment, Infrastructure & Ecosystem
n The Methodology of Tech. Development – Product/Yield Driver!
n Yield (Profitability & Manufacturability) as the Initial Focus, then
Quality and Reliability
n The Learning Curve – Yield Optimization & Iterations
n Tools: Yield & Failure Analyses, H/W, SPC, CPK
n Critical Skill-sets: Root-cause Analyses, DOE (Design-Of-Experiment),
Solution Formulation, STR learning cycles
n Critical Success Factors:
n Baseline, STR, Human, Team, Time, Data, Tools, etc.
Feb. 11, 2015 Technology Development Methodology - CMOS 3
5. The Core Value of CMOS (Patents?!)
CMOS is the “Wheel” of “Information Age”
n Basic switch: the Simplest implementation of “1” or “0”
n The most basic in the Digitization World (IC Designs)
n Signal integrity: the Best, to “Vcc (Power)” & “Gnd” levels
n Power: the Lowest, only switching current (& no DC)
n Structure: the most Fundamental (time invariant)
n 4 terminals: Input, Output, Vcc (Power) and Ground (Zero Ref.)
Feb. 11, 2015 Technology Development Methodology - CMOS 5
Latch-up is the
Achilles’ heel
6. Feb. 11, 2015 6
The CMOS Low Power Advantage (2005)
Picture was
taken on
September 7,
1983 by A.
Kwok (IDT) for
a Fortune
Magazine
article about
IDT (Integrated
Device
Technology)
and CMOS
technology
published on
the Oct. 17,
1983 issue (P.
85) written by
Gene Bylinsky.
IDT went IPO
in Feb. 1984 on
CMOS.
Source: “(Silicon Valley) High Tech – Window to The Future” (Gene Bylinsky)
Technology Development Methodology - CMOS
7. Feb. 11, 2015 7
Source: Intel Presentation
Moore’s Law as the growth engine of hardware platform:
IC cost reduction by 50% in every 18 months or less
CMOS power scaling => Feature size scaling
Feature size scaling => Price scaling
The secret behind scaling: Price can drop
because power can be reduced!
Technology Development Methodology - CMOS
8. Source: Intel Presentation
This future of semiconductor industry
could not happen w/o CMOS!
Feb. 11, 2015 Technology Development Methodology - CMOS 8
9. IDT Founders’ Spin-off (Serial Entrepreneurs)
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Al
Kwok
AMD
AMD
AMD
A/I
IDT
IDT
IDT
IDT
VTI
QSI
QSI
QSI
QSI
QSI
QSI
Self
PMC
P/N
NMI
NMI
NMI
NMI
#256
#
7
#
7
#
10
George
IDT
IDT
IDT
IDT
IDT
IDT
Clon
Laws
Huang
Frank
Lee
IDT
IDT
IDT
IDT
IDT
IDT
IDT
PSI
PSI
PSI
PSI
PSI
PSI
PSI
GVT
GVT
GVT
GVT
GVT
GVT
G/C
Norm
IDT
IDT
IDT
IDT
IDT
IDT
IDT
PSI
PSI
PSI
PSI
PSI
PSI
PSI
PSI
Nova
Nova
NMI
NMI
NMI
NMI
NMI
Godinho
Chun
Chiu
IDT
IDT
IDT
IDT
IDT
IDT
IDT
IDT
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
Q/I
Mano
HP
HP
H/I
IDT
IDT
IDT
IDT
IDT
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
Malwah
Varad
?
?
?
?/I
IDT
IDT
IDT
IDT
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
Nova
NMI
NMI
NMI
NMI
NMI
Srinivasan
Jen
Hong
?
?
?
?/I
IDT
IDT
IDT
IDT
QSI
QSI
QSI
QSI
QSI
QSI
QSI
QSI
?
NY
NY
NY
NY
NY
Feb. 11, 2015 9Technology Development Methodology - CMOS
n Collaborative Innovations leads to Co-founders
n Entrepreneurs create their own jobs
n Serial Entrepreneurs are the major job creators and new
engines of GDP – treasure of a Knowledge-based Society
n IDT (CMOS) & NMI (KBP) enabling IT, impacting >>$10T GDP globally
10. Technology Evolutions Encountered
n IDT (w/ its own fab): Mil. grade; Fastest speed
n CEMOS I (1981-4, 2um): yield improvement only
n Contact mask -> Projection mask, no redundancy, <30% yield
n 100% Wet Etch, Poly gate, 1 layer metal (evaporation)
n CEMOS II (1984-5, 1.5-1.2 um): T.D. & P.D.
n Projection mask, Dry & Wet etch, Silicide gate, Implant, Sputter
n CEMOS III (1986, 1.2-1.0 um): T.D. & P.D.
n Stepper, Dry etch, Contact plug, 2-metal, Ldd…
n BiCEMOS I (1987, 0.8um): 10 ns (T.D. & P.D.)
n Quality Semiconductor Inc. (@ SII & Yamaha)
n Ultra-fast CMOS (1989-90, 0.5um), Idsat ~ 450uA/um
n With back bias (charge-pump ~0.5 mA) – negative feature!
n FCT yielded >95% @ W/S in SII within 1 year (skipped W/S)
n FCT @ F grade; SRAM @ 10 ns (the fastest)
Feb. 11, 2015 Technology Development Methodology - CMOS 10
11. Migration to Low-Power (Battery) Applications
Feb. 11, 2015 Technology Development Methodology - CMOS 11
Portable Battery Operated Devices were dominating since 1990
Low Power IC’s were dominating!
13. The Learning Curve – Core of Tech. Dev.
n In any situation, the Learning Curve starts:
n Where we are: What is the baseline?
n Where should we go: What is the direction for
improvement? What should be the DOE?
n Repeating the process until we achieve our objectives
n Improvements by iterations based on DOE
n DOE must be Well-Conceived (multi-disciplinary!)
n Winner gets more data with less DOE within less time
n Iterative optimization to pin down the global optimal
point(s) of operation w/ least effort (by experimenting)
n There is No Failure in an iterative process
n We Only FAIL, if we GIVE UP (Learning)
n Chinese create their own worst enemy by seeing
iterations (iterative optimizations) as “Failures”
Feb. 11, 2015 Technology Development Methodology - CMOS 13
14. The Big Picture for Technology Development
Lead Customers’ Requirements => 2nd G => 3rd G =>…
Differentiating Value Creations => 2nd G => 3rd G =>…
Business Model => 2nd G => 3rd G =>…
Service/Product Roadmap => 2nd G => 3rd G =>…
Application/Technology Roadmap => 2nd G => 3rd G =>…
IP Portfolio Development => 2nd G => 3rd G =>…
The next innovation ideas come from customers (CRE) & “Stress Tests” identifying
(1) The weak-links in product design & performance and scalability
(2) System (architecture) integrity and scalability (elasticity) issues
(3) The bottlenecks for scalability (technology and manufacturability roadmap)
NetLogic Microsystems is a good benchmark for SMEs as innovators
Close collaboration between the customer (CRE) and the vendor
Mission-&-Time Critical Requirements along Demand-chain
Technology Development Methodology - CMOS 14Feb. 11, 2015
15. Differentiating Value Creation ó Innovation
n Values (to Customers) to get paid - Maximizing
n Performance/Power
n Max. performance at a low (specific) power
n Performance => speed + functionalities (features)
n Battery power for portability, anywhere and anytime
n Personalization
n Quality/Reliability => High Yield too!!
n Food safety assurance is an universal pressing issue
n Delivery => throughput time + logistics + zero waiting
n Just-In-Time + Mini. Process time + Mini. Handling time
n Price => Priced according to values instead of costs
n Price = “Brand (Trust) + Emb. IP + Services” + Cost > 4X Cost
n IP = “Brand (Trademark) + Embedded IP (Patents & Copyrights)
+ Pre-sale Design Win & Post-sale Services (Trade Secrets)”
Feb. 11, 2015 Technology Development Methodology - CMOS 15
16. Product/Tech. Dev. & Intro. Methodology
Feb. 11, 2015 Technology Development Methodology - CMOS 16
New
Technology
Collaborative Innovation
T.D.
17. Differentiate Tech. vs. Product Development
n Product is a “physical” object deliverable to customers
n Technology is the means to make the Product
n Transparent to the customer
n Technology Dev. needs Product Vehicle – Yield Driver
n Ideal Yield Driver facilitates Yield Improvement & Failure Analyses
n Close collaboration among key engineers is paramount
Feb. 11, 2015 Technology Development Methodology - CMOS 17
CEMOS II
Fab Production Transfer
Integration: J. Hong
CEMOS II / 6116X
Product Development
Prod. Engr.: A. Kwok
CEMOS II
Technology Development
T.D. Architect: M. Malwah
Yield/Quality
Reliability
CEMOS II Products
Product Development
Prod. Engr.: Others
Yield Improvement & Failure Analysis Data
18. Technology Development Deliverables
n New Technology Development (process, package
or test) key factors:
n The Process (technology) to make the product
n Process Flow - End-to-end Integration to make the Product
step-by-step, including:
n Process Modules (Specific equipment operations)
n The floor-plan (equipment lay-out) of Production Line
n The Characteristics of the Building Blocks (Elements)
n Electrical Design Rules (EDR) - Device Modeling
n Active components & passive components to make the Products
n For whole Product (IC) simulation (behavior prediction)
n The Arrangements of the Building Blocks
n Topological Design Rules (TDR) - Layout Rules
n How to pack the building blocks closer to make Products – CD’s!
n Such that the behavior modeling (EDR) still valid!
Feb. 11, 2015 Technology Development Methodology - CMOS 18
19. Technology Development
– the Survival Guide:
How to be the Champion?
Feb. 11, 2015 Technology Development Methodology - CMOS 19
20. T.D. Methodology: Increasing Difficulties (I)
n I) Devices (circuit elements) work individually:
n Behave well - perform as expected
n Individual-element Manufacturable - reproducible, spatial invariant
n Reliable - no appreciable degradation due to W/C operational stresses.
n II) Process architecture sound - best case integration:
n Devices can work together to perform (logical) circuit functions.
n Easy topography, wide separation.
n Minimum proximity effects - only allow 1st and 2nd order effects, higher
order effects are excluded.
n Less stringent on equipment performance requirements on consistency,
uniformity, resolution, etc.
n III) Process manufacturability acceptable – W/C integration:
n Circuit functions properly (logics and timings).
n Conform with all topological design rules (including W/C).
n Evaluate lots of proximity effects - higher order effects.
n Equipment meet all design rule performance requirements.
Feb. 11, 2015 Technology Development Methodology - CMOS 20
21. T.D. Methodology: Increasing Difficulties (II)
n IV) Yield/performance optimization & Q/R qualification:
n 1) ≥ 50% wafersort yield (with redundancy) demonstrated
n No systematic yield limiter. Yield is limited by random defects.
n 2) > 90% Pre Burn-in yield (at post packaging process)
n 3) > 95% Post Burn-in yield (after 160 hrs. Vccmax burn-in)
n 4) > 75% Military yield (to 10% Vcc variation @ 125C)
n 5) > 30% yield to the fastest grade (also depends on design)
n 6) Acceptable lifetest (HTOL) results (up to 2000 hrs.)
n 7) No appreciable parametric shifts after 2Khrs. HTOL (trade secret)
n Monitoring Iccsb, Iol, Ioh, taa, Icc, etc.
n 8) Pass all required Groups B & D MIL-STD-883 tests
n 9) Acceptable reliability results on ESD, latch-up, soft-error, hot-electron
effect, etc. (internal requirements)
n 10) Good metal integrity and contact step coverage ≥ 40% (past)
n 11) Tolerance to X-ray (Co-60) radiation ≥ 40 Krad (Si) (prepare for Class S
requirement)
Feb. 11, 2015 Technology Development Methodology - CMOS 21
1988 notes
22. Technology Dev. (IC) Analytical Tools (I)
n I) Fab Process Control / Monitor:
n Prometrix Defect Monitor
n Emission Microscope Analysis Workstation
n Semiconductor Parameter Analyzer (HP 4145)
n Parametric Testers
n II) Reliability Control / Monitor:
n Wafer-level Reliability Monitor.
n Emission Microscope Analysis Workstation
n Semiconductor Parameter Analyzer (HP 4145)
n III) Product Evaluation / Design Verification Tools:
n Dynamical Voltage Contrast and E-beam Prober
n Emission Microscope Analysis Workstation
n Micromanipulator with long working length lenses & hot-chuck
n Laser Line-cutter; Digital and Real-time Oscilloscopes
n Semiconductor Parameter Analyzer (HP 4145)
n Curve Tracer
n Power supplies and Multi-meters
n Function (word) Generator and Logic Analyzer
n Bench Tester.
Feb. 11, 2015 Technology Development Methodology - CMOS 22
23. Technology Dev. (IC) Analytical Tools (II)
n IV) Yield Improvement/ Failure Analysis Tools:
n Wafermap Capability
n Bitmap Capability
n Infra-red Microscope
n Liquid Crystal Hot-spot Detection Kit; Emission Microscope Analysis Workstation
n Semiconductor Parameter Analyzer (HP 4145)
n Failure Analysis Lab:
n Plasma (preferably with Reactive Ion etching capability) Etcher
n Wet Chemical Sink
n Micro-sectioning Station
n Microscope (bright field, dark field and Nomarski)
n SEM:
n Secondary Electron (SE) mode
n Voltage Contrast (VC) mode
n Backscatter Electron (BE) mode
n Electron Beam Induced Current (EBIC) mode
n Dispersive X-ray (EDX & WDX)
n Auger Microscopy
Feb. 11, 2015 Technology Development Methodology - CMOS 23
24. The Baseline (Data) Progression
n Existing Baselines (Reference points for Learning Curves
for yield improvement)
n Success Data Base (wafer lot basis)
n High-low analyses (with Student-t statistics)
n Highest 1/3 vs. Lowest 1/3
n Significant parameters (process variations affecting yields)
n Design-of-Experiment (DOE) - for improving baseline
n Optimization for yields/reliability/speed/power improvements
n STR (Special Test Run) as the vehicle: Super Hot Lots, Hot Lots
n Well-coordinated efforts to expedite, test, analyze and learn STRs
n STR to qualify new process flow, steps, modules and equipment
n New Baseline (B/L) established by at least 2 STRs
n Need at least eq. 1000 hrs lifetest to prove reliability (w/ stress tests)
n Must be very carefully to switch into a new baseline!!
Feb. 11, 2015 Technology Development Methodology - CMOS 24
25. Baseline Data (I)
Feb. 11, 2015 Technology Development Methodology - CMOS 25
Key Information & Data Prd./Prc. Yield Nature of problems
Char. Imprvmt. Systmtc. Random
I Key governing documents:
B/L 1 Electrical Design Rules X X
B/L 2 Topological Design Rules X X
B/L 3 Process (flow) specifications X X
II Design:
B/L 1 Circuit schematics X
B/L 2 Simulation data vs. spec. X Maybe X
B/L 3 Layout plots: composite and single-layer X Maybe X
B/L 4 DRC: violations and marginality X Maybe X
B/L 5 Parasitics extraction / back annotation X Maybe X Maybe
B/L 6 Clock distribution layout X Maybe X Maybe
B/L 7 Data bus routing layout X Maybe X Maybe
III Process:
1 Process device parameters (electrical) X X X
B/L 2 Device paramter trend-charts
3 Process defect monitors (physical) X X Maybe Maybe
B/L 4 Defect monitor trend-charts
5 Lot traveler (overall history) X X X
6 Run cards (at individual process step) X X Maybe Maybe
B/L 7 Station log (status of equipment) X X Maybe Maybe
B/L 8 Equipment SCM trend-charts X X Maybe Maybe
9 Problem log / Eng. disposition record X X X
26. Baseline Data (II)
Feb. 11, 2015 Technology Development Methodology - CMOS 26
Key Information & Data Prd./Prc. Yield Nature of problem
Char. Imprvmt. Systmtc. Random
IV Wafersort / laser repair:
1 W/S test traveler (overall yield summary) X X X X
2 W/S test summary (for individual wafer) X X X X
B/L 3 W/S yield trend-charts (product/process) X Maybe Maybe
4 Wafermaps: composite and single X X X X
5 Bitmaps: composite and single X X X X
B/L 6 Success data base X X X
7 Correlation test procedure/record X X
8 Test hardware log/record X X
V Assembly (packaging):
1 Eng. build inst. (Eng. split lot traceability) X Maybe
B/L 2 Assembly yield trend-charts X Maybe Maybe
VI Class Tests (at Pre B/I, Post B/I and Temp. test):
1 Test Traveler (execution to Eng. inst.) X X X X
2 Test summary (individual split yield data) X X X X
B/L 3 Test yield trend-charts X Maybe Maybe
4 Correlation test record X X
5 Separate into bins (speed grades) X X
6 Reject analyses and correlation X X X
7 AC reject characterization X Maybe X Maybe
8 Speed bin characterization over temp. X Maybe X
27. Root-cause Analysis – e.g., IC Industry
n Observability is essential at all critical steps
n Insert adequate test points along the entire process
n Each test result must reflect specific physical phenomena (the acid test!)
n Capable to pin down to a single variable conclusively
n Overall systematic test points correlation (triangulation for the root-cause)
n Technology ó (Process Flow/mod. + Elect. D/R + Layout D/R) ó Product
n Need to differentiate Technology vs. Product issues
Feb. 11, 2015 Technology Development Methodology - CMOS 27
Wafer
(wafer level)
W/L Process
(Fab.)
Laser Repair
(Redundancy)
Assembly
(Packaging)
Burn-in Vmx
125C 160hrs
Post L/R
Wafersort
Wafer
(die level)
Packaged
unit
Wafer -> Dice
-> Pack’d unit
Pre B/I
Class Test
Post B/I
Class Test
Pre L/R
Wafersort
Elec. Test
Device Para.
Lot Traveller
Run Cards
Wafersort
Summary
Assem. Lot
Summary
B/I Lot
Summary
ProductProductProductProductTechnology
28. Wafersort Flow for Tech Driver (SRAM)
n Open (check continuity = probe card OK)
n @ -0.5V, Imea > 100mA
n Short - All pins except Gnd are at Vcc (CS & OE = Off)
n Two-bit (write A D, read A D, write Abar Dbar, read Abar Dbar)
n Check all decoders, all peripheral circuits
n Gross Functional (Vcc, Vcc/0, 0.5Vcc/0.5Vcc, f > 2 fmax)
n Address Complement March – any failing bit?
n Gross Funct. (Vccmin, Vihmin/Vilmax, Volmax/Volmin) ó Vihmin/
Vilmax & Volmax/Volmin DC tests
n Current measurements:
n Iccmax (Vccmax, Vcc/0, 0.5Vcc/0.5Vcc)
n Isb(CS=Off, Vccmax, dynamic timing)
n Isb2 (CS=Off, Vccmax, static)
n Iin (measure gate leakage)
n Ioh/Iol (optional) – great correlation with Idsat
n Data Retention check – proving each cell can hold data @ power-down
Feb. 11, 2015 Technology Development Methodology - CMOS 28
29. Integrated Solution Formulation (2005)
Key Design Considerations
Noise
(Low)
Performance
(High Speed
& Features)
(Low)
Power
(High)
Density
(/area)
Yield/
Reliability
(High)
Form Factor/
Thermal Stress
(Small & Robust)
Reducing
Overall
Cost to End
Customers
SystemInterfaceFactors
FabTechnologyFactors
Package Technology
Feb. 11, 2015 Technology Development Methodology - CMOS 29
"Design-in Methodology” for "Doing the Right Things Right at the First Time" mindset
(to proactively solve problems at their sources of creation) to enable "First-Silicon”
Success: Marketability, Scalability, Robustness, Manufacturability, Reusability, Quality/
Reliability, Testability, etc. – All for “Integrated Design Capability”
Costs to Society
is the Wildcard
30. A Process Control System
Feb. 11, 2015 Technology Development Methodology - CMOS 30
31. Process Optimization and Control
n Process Integration
n Device architecture and engineering to EDR & TDR
n Process Modular Development
n Masking, etches, implants, diffusions, thin films, interconnects
n Process Development Methodology
n Process characterization
n Actual vs. Stimulation; Build Success Data Base
n Process (equipment) capability determination
n CD variation vs. yield: Find yield cliffs for equipment controls
n Process optimization
n Optimization of yields/speed/power and process windows
n Process Control
n 100% Critical Variables with Cpk > 1.5 & Cp > 2.0
Feb. 11, 2015 Technology Development Methodology - CMOS 31
32. STR (Special Test Run) Evaluation
Roadmap & Methodology
Feb. 11, 2015 Technology Development Methodology - CMOS 32
33. STR Evaluation Roadmap
Feb. 11, 2015 Technology Development Methodology - CMOS 33
n Test flow is designed to
pinpoint problem a
variation step at a time
n Non-destructive stress
tests are needed for
early warning system
n The worst problem is
the walking wounded
causing reliability
problem – need to
screen them out
n Quality overrides yield –
better to reject good
than to accept failure
34. STR Evaluation Methodology (I)
n I) Process Evaluation:
The following information must be collected systematically and diligently by Fab
Engineering and/or Technology Development and made easily accessible:
n Post metal electrical data: E.T. (Electrical Tests), defect monitor data,
etc.
n In-line process control data: CD measurements, thickness
measurements, sheet resistance measurements, etc.
n Run problem logs: Engineering Hold Cards, Q.C. Inspection Records
and Run History Card, Sort Alert Card, etc.
n Experiment records: Engineering Split Information, STR (Special Test
Run) Description, STR data (measurements, photos and SEM), etc.
n II) Wafersort:
The following practices are imperative for STR evaluation or yield improvement:
n A) Yield data - wafersort result: The individual test must be well
designed to both:
n 1) Verify if the (gross) product specifications are met
n 2) Monitor specific device, process or design problem
Feb. 11, 2015 Technology Development Methodology - CMOS 34
35. STR Evaluation Methodology (II)
n B) Yield analyses:
n 1) Yield comparison to the Success Data Base (SDB) as the baseline reference.
n 2) Statistical analysis (mean and sigma calculation) on key W/S yield/failure, PCM
(device & defect) and process parameters (in-line process control data)
n 3) High/low (yielding wafer) data comparison analysis to determine the key
cause(s) for yield variation within the lot by:
n a) Separate the highest and lowest yielding wafers into 2 groups -- minimum 7 wafers
each.
n b) Apply Student-t statistics to identify the key parameters that affect yield most.
n c) Proceed with wafermap and bitmap comparison analyses if needed.
n C) Wafermap analysis:
n 1) Generate a map on the wafer sorted to indicate the W/S yield or failure
category of an individual die.
n 2) Generally, there are 2 kinds of wafermaps:
n a) Single wafermap of an individual die.
n b) Composite wafermap of a group of wafers on a single (or a set of) yield or failure
categories.
n 3) Yield/failure pattern study can show if yield is limited by any systematic
problem: e.g., a repeating failing die can indicate mask or masking defects. This
is a necessary check to qualify new masks.
n 4) High/low analysis with wafermap is a very powerful tool.
n Feb. 11, 2015 Technology Development Methodology - CMOS 35
36. STR Evaluation Methodology (III)
n D) Bitmap analysis (for memory products only):
n 1) Generate a map (according to address and data locations) of failing
bits (to a specific functional test/condition) on an individual die.
n 2) Generally, there are 2 kinds of bitmaps:
n a) Single bitmap of an individual die.
n b) Composite bitmap of a group of dice.
n 3) Bitmap pattern study can show if yield is limited by any systematic
problem: e.g., a (or a set of) repeating failing bit(s) can indicate mask
or masking defects. This is a necessary check to qualify new masks.
n 4) Failing bits are to be categorized according to specific "patterns":
column, row, pair, cluster, line, etc.
n 5) Each failing bit pattern category is to be analyzed with Strip-back
visual/SEM inspection.
n 6) High/low analysis with bitmap is a very powerful tool.
n 7) Defect density can be estimated by counting and visual identification
of the failing bit patterns.
Feb. 11, 2015 Technology Development Methodology - CMOS 36
37. STR Evaluation Methodology (IV)
n E) Strip-back visual inspection:
n 1) Select the sample (representing specific failing category/
pattern) from wafermap and bitmap analyses
n 2) Use wafermap and bitmap to locate the failing memory cell
and see if the defect is visible.
n 3) Etch back the wafer layer-by-layer one at a time to identify
the defect (with high resolution microscope and SEM) with
comparison to the adjacent good cell.
n 4) The etch-back should be well controlled with proper wet
chemicals or plasma etch. Sometimes, staining is needed to
"highlight" the defect -- e.g., silicon pits.
n 5) Make sure no artifacts are introduced in the process.
n 6) Identified defects should be categorized w.r.t. the failing bit
patterns for consistency check and future reference.
Feb. 11, 2015 Technology Development Methodology - CMOS 37
38. STR Evaluation Methodology (V)
n F) Integrated data cross-check and further investigation:
n 1) All data (including the process data) must be mutually consistent (not
contradictory) and supportive.
n 2) The sources of major yield problems must be identified and fixed, whether
they are related people (e.g., human errors), equipment, materials, methods
(e.g., specification problems), or environment.
n H) Case record publishing and filing:
n 1) The findings must be communicated, discussed and digested for:
n a) prevention of recurrence -- top priority
n b) assessment of the impact on all affected products -- for proper containment and
disposition.
n c) future process/design improvements.
n 2) The case record with all data and photos must be filed properly - they are the
treasure of the company.
n I) Comments:
n The investigation and analyses at wafersort are inter-related and sequential --
the previous one followed by the next one,
n E.g., the hi/lo yield analysis guides the wafermap analysis, the wafermap
analysis guides the bitmap analysis, the bitmap analysis guides the strip-back
visual inspection and so on.
Feb. 11, 2015 Technology Development Methodology - CMOS 38
39. Closing Remarks
n Technology Development Leadership is indicative of a well-
coordinated learning organization progressing with solid baselines and
well-conceived iterative optimizations to achieve maximum results with
minimum efforts within the shortest time.
n Collaborative Innovation is paramount as the product goes through
various stages of different domain expertise led by different people,
the project hand-offs from one stage to the next have to be seamless
and smooth for achieving the fastest time-to-market & time-to-mass-
production.
n Design-in Manufacturability and Testability have been the critical
success factors to differentiate the market leader.
n CMOS early innovators (IDT & Cypress) did not use patent litigations
as entrance barriers to impede their competitors, instead industry-wide
collaboration and knowledge sharing along the supply chain with
equipment suppliers facilitated and expedited ecosystem building for
CMOS to be the Mainstream IC Manufacturing Standard
Technology (by 1986).
Feb. 11, 2015 Technology Development Methodology - CMOS 39