Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • How to Use This Course
  • Reading Path
  • STA Glossary
  • STA Metrics Reference
  • STA Mastery Cheatsheet
  • Timing Paths and Checks16
  • Constraints & SDC Fundamentals20
  • Delay Graph & Path Analysis16
  • Beyond Setup and Hold16
  • Multi-Corner Multi-Mode Signoff16
  • OCV, AOCV, POCV and Variation16
    • OCV, AOCV and POCV Mechanisms
    • Derate Policy and Approval
    • Path Depth Sensitivity
    • Variation Signoff Dashboard
      • What to Track
      • Margin Trends
      • Comparing Runs Honestly
      • Signoff Package
  • Crosstalk and Noise16
  • ECO and Closure16
  • STA Debug Playbook16
  • STA Interview Prep26
  • Capstone Timing Signoff Lab4
Loading lesson

‹Variation Signoff Dashboard
  1. Topics
  2. Static Timing Analysis
  3. OCV, AOCV, POCV and Variation
  4. Variation Signoff Dashboard
  5. Signoff Package

Part 6 · OCV, AOCV, POCV and Variation · All levels

Signoff Package

A variation signoff package states the method, data, margins, results by scenario, exposure and open items so a reviewer can check the claim.

On this page

    Why it matters

    A signoff claim is checked by someone who was not in the closure meetings. The package is everything that person needs: what was run, with what data, what the results were, what is still open. A package that is missing an item cannot be accepted, whatever the numbers say.

    Contents

    Item

    What it shows

    Method and data versions

    OCV method, tables or sigma data, library and extraction versions

    Margin ledger

    Every margin with owner and evidence

    Results per scenario and union

    WNS, TNS, violations, with the near-critical histogram

    Derate audit output

    Applied values against policy

    Comparison with the previous milestone

    On common endpoints, with added and removed ones

    Open items and waivers

    Each with an owner and a date

    example.py
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    required = ["method and data versions", "margin ledger", "results per scenario", "near-critical histogram",
                "derate audit", "milestone comparison", "open items"]
    submitted = {"method and data versions": True, "margin ledger": True, "results per scenario": True,
                 "near-critical histogram": True, "derate audit": False, "milestone comparison": True, "open items": True}
    missing = [r for r in required if not submitted.get(r)]
    wns_union = 0.012
    numbers_clean = wns_union >= 0
    accepted = numbers_clean and not missing
    assert missing == ["derate audit"]
    assert accepted is False
    print("PASS: package not accepted; missing", missing, "although union WNS is", wns_union)

    The union WNS is positive (+0.012 ns), but the package lacks the derate audit, so nobody has confirmed that the derates that produced the number are the policy ones. Acceptance depends on completeness first and numbers second: a clean number from an unaudited setup proves nothing about the setup.

    What the review looks for

    • Every item above is present, dated and linked to the run that produced it.

    • A reviewer who was not involved can reproduce the headline numbers from the package alone.

    • Open items have owners; waivers expire.

    Common pitfall

    Assembling the package from the latest run only. The comparison with the previous milestone and the margin ledger history are what show the claim is not an accident of one run.

    Interview

    Describe the package as the evidence trail, not a report.

    Check yourself Why is the derate audit part of the package even when the numbers pass?

    It confirms the applied derates match the policy; a missing or wrong derate makes passing numbers meaningless.

    Related topics

    • Comparing Runs Honestly

    Previous

    105 / 183

    Next: Crosstalk and Noise

    Up next

    Up nextCrosstalk and Noise

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • What to Track
    • Margin Trends