Checklists turn the gate from a meeting into a procedure. Each owner answers only their own questions, in writing, before the review — so the meeting spends its time on disagreements rather than discoveries. The three lists below are the minimum; projects add rows, never remove them without recording why.
Owner | Checks |
|---|---|
Design | Every crossing has an intended synchronising structure; sources are registered with no logic before the first flop; protocols (handshake, Gray, FIFO) match the data type; boundary crossings name their owning side |
Verification | Protocol assertions are written and passing; metastability-injection checks cover the synchronisers; quasi-static assumptions have assertions that re-check every regression; coverage includes stall and back-pressure behaviour |
Signoff | Setup reviewed against the architecture; structural runs cover the full hierarchy at the signed-off revisions; every finding is fixed or waived with evidence; constraints reviewed against the inventory; dashboard history is continuous across revisions |
Each owner completes their list in writing before the review and attaches the evidence references.
Items that fail become defects or waiver requests with owners and dates — never silent acceptances.
Disagreements between lists (design claims a protocol the verification list cannot find) are the review agenda.
The completed lists are stored with the signoff record, so a later question about any crossing has a written answer.
The reset crossing rows deserve a cross-check with the reset material: reset strategies for synchroniser structure and reset domain crossing for the failure modes the checklists guard against.
Both: design's structure claim lacks verification's proof, and verification's list lacks the assertions. The review records a defect — write the assertions or waive with other evidence — with an owner and a date.
Because signoff is a claim about evidence, and the lists are the index to that evidence. A later ECO, audit, or failure analysis needs the written answers, not memories of the meeting.