Professional Documents
Culture Documents
Ravikumar
Presenters
2
Presenters: 1.
2.
3.
Outline - Day 1
3
(9:00 am 10:00 am) (10:00 am 10:45 am) (11:00 am 12:30 pm) (2:00 pm 2:30 pm) (2:30 pm 4:00 pm) (4:15 pm 5:00 pm)
Outline - Day 2
4
(9:00 am 10:00 am) (10:00 am 10:45 am) (11:00 am 11:15 am) (11:15 am 12:30 pm)
Outline - Day 3
5
Test Compression
Coffee Break
Fault Diagnosis
Lunch
Industrial Experiences
Acknowledgements
6
Xioagang Du, Mentor Graphics Dhiraj Goswami, Mentor Graphics Ruifeng Guo, Intel Xijiang Lin, Mentor Graphics Srinivas Patil, Intel Janusz Rajski, Mentor Graphics Nagesh Tamarapalli, Mentor Graphics Nandu Tendolkar, Freescale Jerzy Tyszer, Poznan Institute of Technology
Acknowledgements
7
Sudhakar Reddy acknowledges the help of University of Iowa graduate students Zhuo Zhang, Santiago Remersaro, Chen Liu, Narendra Devtaprasanna
Outline - Day 1
8
(9:00 am 10:00 am) (10:00 am 10:45 am) (11:15 am 12:30 pm) (2:00 pm 2:45 pm) (2:45 pm 4:00 pm) (4:30 pm 5:30 pm)
Outline
10
Design Characteristics Defects on DSM Circuits Systematic vs. Random Defects Objectives of Digital Testing Defect Coverage, Yield, Test Quality Functional vs. Structural Tests Benefits of Structural Tests Factors Affecting Test
180 nm 90 nm
R
R = 50 K Delay of 250ps How do you test for it?
Year Node [nm] Devices [million tr.] Voltage [V] Oxide thickness [nm] Static power per dev
70nm
System-on-a-Chip Characteristics
13
System architecture
Microprocessors, DSP cores Buses, peripherals, memory ASIC portion
CPU core
ASIC
Memory
ASIC
Memory
Structures: Logic, memory, analog Multiple embedded memories: SRAM, DRAM, Flash, CAM Analog and mixed signal: PLLs, clock recovery Field programmable logic RF cores: wireless receivers IP cores and reusable blocks available from multiple vendors Design efficiency achieved by hierarchical core-based design style
I/0
IP core
PLL
100 million gates 400 million transistors Billions of vias 6 km of wire / cm2
Al 4-5 Levels
A Transistor
?
Defect distribution change with process
Cu (8 Levels)
Low-K Dielectric
Cu Plugs
A Transistor
Resistive shorts
Vias
Quality Requirements
18
Escapes
1-p
nt me hip s
Faults detected
1-Y
Quality Requirements
19 Escapes = (1 - Y)(1 - p)
0.01 0.009 0.008 0.007 0.006 0.005 0.004 0.003 0.002 0.001 0
0.990 0.991 0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999 1.000
Yield = 0.9
Yield = 0.1
What is Yield?
20
Fabrication Electrical Test Fab Metrology Wafer Sort Assembly Final Test
1M.
Rencher and G. Allan,GDSII Yield Signoff Methods, Design for Yield, EE Times
Design Re-spins
22
100 90 80 70 Rate (%) 60 50 40 30 20 10 0 350 250 180 Technology Node (nm)
Source: Mark Miller, eeDesign.com, 2004
130
90
Target
Feature related defects are becoming more prevalent than particle-driven defects in nanometer designs1 Better and more effective (Higher Quality) testing is necessary to maintain quality (or DPM) levels
1 Hsu and Chen TSMC: Scan Diagnostics in the Nanometer Design Era; Semiconductor Manufacturing, March 2006.
DPM is a function of yield and test quality As yield decreases, DPM increases Highest test quality is needed just to maintain DPM levels
Functional tests are tests developed to mimic functional behavior of the circuit Disadvantages
Requires detailed circuit knowledge Expensive development and troubleshooting Difficult to create high coverage patterns Complicated failure diagnosis Often requires high performance testers Large inter-dependent test sets
Targeted for detecting manufacturing defects in the device Use fault models to enable test generation
Correlates to defects Proven in industry Predictable results Provides high coverage tests, quickly
What is DFT?
29
Inserts structures inside the IC to make it easier to test Generates test sets Provides measures of test quality Diagnoses the cause of a failing device
Test quality impacts your product quality Assume 90% yield and 1M chips shipped/yr
99% defect coverage => 1053 defective chips are shipped 95% defect coverage => 5254 defective chips are shipped Includes coverage for all defects: at-speed, bridging, stuck-open, etc.
Shortens Time-to-Market
31
Time-to-market pressures
More test (e.g. at-speed) Higher coverage Bigger designs Longer runtimes
DFT tools must keep pace with design technologies and test methodologies
Reduces ATE memory requirements Reduces test time Minimizes or eliminates tester reloads Improves tester throughput
Coverage
Test Cost
Escapes
Pattern Volume
VDD
IDDQ Testing
34
Models high current flow through a gate Measures quiescent power supply current during the stable state Involves measuring current - not logic values
VDD
A B 0 0/1 Y P1 1/ 0 0
A B
P2
Y
1/0
VSS
IDDQ Testing
35
Iddq
1000
PolySi PolySi
PolySi
100
10
Silicon
Silicon
1 130 90 65 45 32 22
Technology nodes
ASIC 1 2 3 4 5
At .18m 1% defects were speed-related At .13m 2% defects were speed-related 30-70% DPM reduction w/at-speed test
* Effectiveness Comparisons of Outlier Screening Methods for Frequency Dependent Defects on Complex ASICs VTS 2003. ** Delay Defect Characteristics and Testing Strategies IEEE Design & Test of Computers, Sept-Oct 2003.
Phases of Test
38
Fabrication Fabrication
Slice & Dice, Package Slice & Dice, Package Burn-in, Burn-in, Packaged Part Test Packaged Part Test Ship to Customer Ship to Customer
DC Parametric Test logic verification DC logic stuck-at AC frequency assessment AC logic delay Memory testing Memory retention IDDQ test Specialty vectors
Scan configuration Test pattern generation Test coverage Test application time Cost of test Diagnosis
Conclusions
41
Better and more effective testing is necessary to maintain quality (or DPM) levels
Scan test and diagnostics are playing an increasingly important role in failure analysis and yield learning