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
  • Crosstalk and Noise16
  • ECO and Closure16
  • STA Debug Playbook16
    • Constraint Bug Patterns
    • View Mismatch Triage
    • Path Group War Room
    • Signoff Regression Hygiene
      • Run Comparison
      • Constraint Diffing
      • Reproducibility
      • Final Checklists
  • STA Interview Prep26
  • Capstone Timing Signoff Lab4
Loading lesson

‹Signoff Regression Hygiene
  1. Topics
  2. Static Timing Analysis
  3. STA Debug Playbook
  4. Signoff Regression Hygiene
  5. Final Checklists

Part 9 · Debug Playbook · All levels

Final Checklists

A final checklist ranks items as blockers or majors and gates signoff on blockers closed and majors waived with evidence.

On this page

    Why it matters

    By the end of a project there are many things to confirm, and they differ in weight. A checklist that treats a missing clock definition and a missing log file alike teaches nobody anything. Weighting items into blockers (must be closed) and majors (closed or waived with evidence) gives a decision rule.

    The checklist

    Item

    Weight

    Evidence

    No unclocked registers or unconstrained endpoints

    Blocker

    Coverage report is empty

    Every exception reviewed with its path count

    Blocker

    Exception register

    All scenarios run, setup and hold clean

    Blocker

    Per-scenario summary

    Multicycle setup exceptions have hold counterparts

    Blocker

    Exception register

    Derate audit and library sanity checks passed

    Blocker

    Audit output

    Run manifest complete and reproducible within tolerance

    Blocker

    Manifest and comparison

    SI and noise: critical receivers and filters clean

    Blocker

    SI checklist

    SI and noise signoff: remaining items

    Major

    SI checklist

    Waivers documented with owner and expiry

    Major

    Waiver list

    example.py
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    items = [
        ("coverage", "blocker", "closed"),
        ("exceptions reviewed", "blocker", "closed"),
        ("all scenarios clean", "blocker", "closed"),
        ("multicycle hold counterparts", "blocker", "closed"),
        ("derate audit and sanity checks", "blocker", "closed"),
        ("manifest reproducible", "blocker", "closed"),
        ("si/noise critical receivers and filters", "blocker", "closed"),
        ("si and noise signoff (remaining)", "major", "open"),
        ("waivers documented", "major", "closed"),
    ]
    blockers_open = [n for n, w, s in items if w == "blocker" and s != "closed"]
    majors_open = [n for n, w, s in items if w == "major" and s == "open"]
    majors_waived = [n for n, w, s in items if w == "major" and s == "waived"]
    ready = not blockers_open and not majors_open
    assert blockers_open == [] and majors_open == ["si and noise signoff (remaining)"] and majors_waived == []
    assert ready is False
    print("PASS: not ready; open majors", majors_open, "; waived majors", majors_waived)

    All seven blockers are closed, including the derate and library sanity audit, reproducibility against the manifest, and the critical SI and noise receivers and filters: nondeterminism is a defect to fix first, open critical SI and noise items block signoff, and a failed library sanity check is not waivable. One major is still open: the remaining SI and noise items. The rule says that signoff waits for the open major to be closed or formally waived. Showing the waived items separately keeps accepted risk visible: a waived item is not a closed item, and the signoff statement lists it. The weights here are an example policy; adjust them to your project, but keep reproducibility, the library and derate audit, and critical SI and noise receivers as blockers.

    What the review looks for

    • Each item has a weight, an owner, an evidence link and a status of closed, waived or open.

    • Waived items carry the waiver text, approver and expiry date.

    • The checklist is regenerated from the current state, not carried over from the previous milestone.

    Common pitfall

    Letting 'majors' be quietly dropped. A major that is neither closed nor formally waived is an open risk with no owner.

    Interview

    Describe the rule: blockers closed, majors closed or waived with evidence.

    Check yourself If 'all scenarios clean' were open, could signoff proceed with a waiver?

    No: it is a blocker. Blockers have to be closed; only majors can be waived.

    Related topics

    • Reproducibility

    Previous

    153 / 183

    Next: STA Interview Prep

    Up next

    Up nextSTA Interview Prep

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • Run Comparison
    • Constraint Diffing