Each question below is asked as a review disagrees with you. The strong answer takes a position, states the strongest counter, and names the evidence that would flip it. That third element is what interviewers score.
Field | Answer |
|---|---|
Interviewer is testing | Prioritisation. |
Bad answer | "Review all 200 equally." |
Better answer | "Triage by consequence: control and reset crossings first, then buses, then static config; sample, do not skim." |
Strong answer (a 4) | Adds the method (crossing inventory against intent, violation classes first, reconvergence and reset ordering before synchroniser legality), prices sampling versus coverage, and names what flips it (a prior escape in the deprioritised class). |
Follow-ups | What you delegate; what the dashboard must show daily. |
Common trap | "Trust the clean report". |
Evidence to cite | The triaged review log. |
Taught in |
Field | Answer |
|---|---|
Interviewer is testing | Latency-versus-margin reasoning. |
Bad answer | "Two stages is the rule, no exceptions." |
Better answer | "Stages buy settling time against the failure rate; two is the minimum, and a cut from three to two needs per-crossing rate, activity, and consequence analysis, not a blanket change." |
Strong answer (a 4) | Treats two stages as the non-negotiable minimum (a single flop gives essentially no settling time and is never signoff-legal), prices the latency saved against the margin lost for any proposal to go from three stages to two on high-frequency or high-reliability crossings, and names what flips it (measured margins with headroom, rate and consequence analysis, owner sign-off per crossing). |
Follow-ups | Which crossings you would never cut; third-stage economics. |
Common trap | Blanket rules in either direction. |
Evidence to cite | Per-crossing analysis, never a global dispensation. |
Taught in |
Field | Answer |
|---|---|
Interviewer is testing | Boundary ownership. |
Bad answer | "Trust the vendor." |
Better answer | "The crossing list, the synchroniser per crossing, clock and reset intent, and the verification evidence, before integration." |
Strong answer (a 4) | Adds the containment (treat all boundaries as unsynchronised until proven, re-verify at chip level, own the assumptions in writing) and the schedule consequence (integration waits for intent). |
Follow-ups | What you check yourself versus accept; black-box structural assumptions. |
Common trap | "It worked in the last chip". |
Evidence to cite | The delivered intent plus chip-level re-analysis. |
Taught in |
Field | Answer |
|---|---|
Interviewer is testing | Cross-discipline reasoning. |
Bad answer | "Gating does not affect CDC." |
Better answer | "Every gated clock is a domain boundary event: crossings into and out of gated regions need analysis with the gate modelled, and enable timing matters." |
Strong answer (a 4) | Names the mechanisms (lost events into stopped clocks, skew on re-enable, reset interaction with gated domains), demands the gated-clock inventory as intent, and prices power against verification load. |
Follow-ups | Clock mux versus gate; enable synchronisation. |
Common trap | Analysing with gates tied open. |
Evidence to cite | Analysis with gating modelled plus enable-timing checks. |
Taught in |
Field | Answer |
|---|---|
Interviewer is testing | ECO discipline. |
Bad answer | "Placement does not change logic." |
Better answer | "It changes skew, settling time, and possibly legality: re-run structural analysis, reconvergence, and the affected proofs on the post-ECO netlist." |
Strong answer (a 4) | Scopes the re-verification (affected clock pairs, reconvergence cones, reset branches), names the gate-level concern (delay shifts on synchroniser inputs), and refuses signoff carry-over across the ECO. |
Follow-ups | What the ECO report must list; spare-flop placement. |
Common trap | "Functionally equivalent, so still signed off". |
Evidence to cite | Post-ECO re-analysis, not the pre-ECO package. |
Taught in |
Field | Answer |
|---|---|
Interviewer is testing | The whole course in one answer. |
Bad answer | "The waivers are low severity." |
Strong answer (a 4) | What a 4 sounds like: "Assumption: each waiver is scoped to stated RTL, intent, and constraints. Mechanism per waiver: state the failure each one could cause and why the evidence rules it out. Trade-off: the structural alternative for each, priced, and why the waiver was proportionate. Evidence: per waiver, owner, condition, expiry, attached proof, and the re-analysis date. Residual risk in one honest paragraph. What flips my position: any ECO touching the scoped logic, any intent change, or expiry, each triggering automatic re-review. I am not asking you to trust the waivers; I am asking you to audit this package." |
Follow-ups | Pick the weakest of the three and attack it; expiry mechanics. |
Common trap | Defending waivers as a group instead of one by one. |
Evidence to cite | The auditable package. |
Taught in |
The flip condition: what evidence would change the verdict. A position without a flip condition is a preference; with one, it is engineering judgement.