Title: DFT Technologies for HighQuality LowCost Manufacturing Tests
1DFT Technologies for High-Quality Low-Cost
Manufacturing Tests
- Yuval Snir
- Advanced Digital design flow seminar 2006
2Agenda
- Background facing with new era of defects
- Concepts Models
- Problem Regards testing with low cost
- fast technologies
- Main goal Two effective implementation
- solutions
3Background
- The Shrinkage of semiconductor devices brought a
new distribution of defects - Old methods such as the stuck-at-fault-model and
standard memory BIST are no longer adequate. - Timing defects have become more significant (at
least 2 of all defects) - Need higher quality tests for reducing the
defective chips (DPM) -
4Background
- For 130 nm fabrication
- Of yield 70
- No statistical faults
- Defect rate will be 0.7
- 7000 defective devices
- un detected per million
- (DPM)
-
5Concepts Models
- ATPG Automatic Test-Pattern Generation
- On die hardware tool that generates patterns for
false detection. - Test-Patterns Behavior of output as a function
- Of the inputs
- Scan Chains vectors of inputs/outputs
6Concepts Models
- Stuck-at fault-model
- the most popular fault model used in practice
- a line whose status is stuck at a given value
(normally 0 or 1). - Input 1,1 reveals S_at_0
- fault
7Concepts Models
- At-speed fault model
- Test for timing faults
- most companies use this test today
- Check the duration in which a logic gate Change
its output - Check whether a circuit behave as expected at the
right speed typically high clock frequencies - The most common detection is the broadside
transition pattern which checks for slow-to-rise
slow-to-fall faults. - Broadside patterns involve loading scan cells to
produce both an initial value and an opposite
value ready to propagate through the path.
8Concepts Models
- At-speed fault model
- Example
- Initialization
- B1
- Opposite value
- Propagates from
- B to C
capture
launch
9Concepts Models
- At-speed fault model
- Two possibilities to
- Perform the at-speed
- test
broadside
launch of shift
10The Problem
- Desired fast low cost technologies tests
- Transition pattern set is 3-5 bigger then
- stuck-at pattern set
- Sometimes there isnt enough room on the tester
memory for both pattern sets. - Yields expensive tester reloads of the memory.
11solutions
-
- 1. Effective Merging At-Speed with Stuck-At
Patterns - 2. EDT Embedded Deterministic Test
12Merging At-Speed with Stuck-At Patterns Sets
- The TDF (Transition delay fault) patterns also
detect a significant percentage of stuck-at
faults - Truncate TDF patterns
13Merging Patterns - Case Study
Design demands
Characteristic of the design
- The tester can hold up to 10,000 test pattern
- The highest priority best possible coverage for
stack_at_ - One TDF pattern set for each clock domain and one
- For cross clock domain.
14Merging Patterns - Case Study
Test generation results before optimization
- Due to the typically slow operation of the tester
- The TDF test coverage is only 85.14.
15Merging Patterns - Case Study
Test generation results before truncation
optimization
16Merging Patterns - Case Study
- Flow for generating higher coverage
- Arrange TDF test patterns from most significant
to least - Truncate TDF patterns ( 90 of the overall
achievable) - Fault grade the truncated TDF patterns for SAF
- Generate top-off SAF pattern set
17Merging Patterns - Case Study
Test generation results after truncation
optimization
18Merging Patterns - Case Study
Test generation methodology comparison
19Review -The Problem
20EDT-Embedded deterministic test
An Embedded deterministic test architecture
21EDT-Embedded deterministic test
- Minimize test time and utilize memory
- Uses embedded compression logic
- Enables scan chains to increase by up to 100x
- Not necessitate special design
- Bypass-Test generation can be done also in ATPG
22EDT-How this is work?
- With ATPG 1-5 of scan cells get specified.
- With EDT there is a compression of specified
cells -
23