( ESNUG 491 Item 10 ) ------------------------------------------- [05/12/11]
Subject: The correct apples-to-apples Mentor TestKompress compression data
> However, I'm not sure I trust the results. Typically, fault coverage is
> roughly the same or slightly less *with* scan compression, compared to
> the non-compressed run. The results here are opposite; the compressed
> fault coverage is 90.55% while the non-compressed run is 84.89%, which
> is significantly less (~5%).
>
> - [ Doubting Thomas ] from ESNUG 491 #5
From: Debo Sekoni <dsekoni=user domain=gmail not mom>
Hi John,
I'd like to apologize for the discrepancy in our reported test coverages.
At the time the report was generated, our focus was on integrating Magma
Talus Design with Mentor Tessent TestKompress. Our intent was not to
manipulate our coverage results, rather to share with other engineers this
new flow.
The reason for the discrepancy lies in the fact that, the Non-Compression
results were based on a post-scan-inserted, timing-closed, *post-layout*,
GDSII netlist database with buffers, spare cells, and spare flops.
Whereas the Compression results were based on the post-scan-inserted,
*pre-layout* netlist (*from the same database*). The keyword here is
post-layout versus pre-layout.
The test coverage matrix below compares the metrics for the Non-Compressed
patterns to the Compressed *based* on the number of output loads only
(apples-to-apples):
Non-Compression Scan Compression
(post-scan, pre-layout)
Test Coverage: 92.68% 92.33%
Fault Coverage: 90.58% 90.16%
ATPG Effectiveness: 99.87% 99.85%
Chain test Patterns: 111,348 b 169,488 b
Transition test Patterns: 2,585,945,952 b 26,750,856 b
Stuck-AT test Patterns: 337,384,440 b 4,135,122 b
================= =============
Total Scan Test Patterns: 2,923,441,740 b 32,264,352 b
Total Scan Test Pattern Volume: 2.72 Gb 0.030 Gb
I apologize for this discrepancy and would like to thank "Doubting Thomas"
for his pointing this out.
- Debo Sekoni
DFT/DFM Contractor
Join
Index
|
|
