Extended case study

Release review: diagnosis resolution, top failing buckets, yield learning turnaround regresses after a test-flow update touching Diagnosis & Yield Learning.

Background

Team had prior signoff, then a new program/config introduced regressions in selected buckets.

Symptoms observed

• diagnosis resolution, top failing buckets, yield learning turnaround regression

• Mismatch between simulation and tester

• Escalation without clear owner

Investigation timeline

1. Hour 0: freeze pattern set, constraints, and tester program tags

2. Hour 1: isolate first failing bucket by mode/lot

3. Hour 2: verify legality and constraints assumptions

4. Hour 3: correlate with physical/timing/power context

5. Hour 4: choose minimal reversible fix

6. Hour 5: run full signoff regression matrix

7. Hour 6: publish decision memo and owners

Root cause

Root cause tied to Diagnosis & Yield Learning: Fail logs and diagnosis map failing signatures back to likely defect sites, enabling systematic yield learning loops with design and process teams.

Fix and validation

• Apply bounded fix with owner

• Re-run diagnosis report, fail-log bucket table, yield pareto dashboard

• Re-validate quality, timing, and test power

Lessons learned

• Tag every run artifact

• Mechanism first, command second

• Close with explicit release decision

CASE STUDY - Diagnosis & Yield Learning
baseline metric / regressed metric / post-fix metric

Sequence under stress

DFT FLOW - Diagnosis & Yield Learning

scan insertion -> chain stitch -> compression map -> ATPG -> tester apply -> diagnosis
      |                |                |            |             |
 controllability   shift balance    channel use   coverage     silicon correlation

Primary metric: diagnosis resolution, top failing buckets, yield learning turnaround

DFT deep dive

Compression saves tester time only when diagnosis observability remains credible.

Concept diagram

COMPRESSION LOOP

EDT/decompressor -> compressed patterns -> compactor responses -> diagnosis

Metric graph

PATTERN vs COVERAGE

coverage up   -> pattern count up
compression up -> pattern count down (until aliasing risk)

Reports and artifacts

• compression ratio dashboard

• pattern count trend

• X-source report

• diagnosis bucket summary

Mini case study

Compactor overflow plus unknown inflation caused false diagnosis; masking policy and channel map corrected.

Debug branches

• Separate X issues from silicon defects

• Replay failing patterns uncompressed

• Track tester memory budget

Senior review question

Ask: what evidence proves this DFT decision is safe for production?

Key takeaways

• State metric, lot/corner context, and pattern tag with every claim.

• Treat timing, power, and quality as one signoff problem.

Common pitfalls

• Chasing coverage without legality checks.

• Ignoring test-power side effects of pattern changes.

• Debugging silicon without reproducible tags.

Principal DFT review addendum

Fail logs and diagnosis map failing signatures back to likely defect sites, enabling systematic yield learning loops with design and process teams.

Metric: diagnosis resolution, top failing buckets, yield learning turnaround