Construct a scenario where the back-annotation precedence order produces a subtly wrong result that passes all obvious checks.
From PDVerse PrimeTime STA Interview Guide · pdVerse Mentor Guide
Ten-second Interview Answer
A forgotten set_resistance 5 [get_nets long_route] is tier 2 and outranks tier-3 detailed parasitics, so PrimeTime uses the stale lumped 5 ohms while report_annotated_parasitics -check still passes because the SPEF data is present, just shadowed.
Complete Technical Explanation
Picture this. Months ago someone added set_resistance 5 [get_nets long_route] to a shared constraint file to model a route before extraction existed, and nobody removed it. Later the team reads an accurate full-chip SPEF and runs report_annotated_parasitics -check, which passes - the net does have a complete RC network from SPEF. Everyone believes long_route uses the extracted resistance. But set_resistance sits at tier 2 and outranks the tier-3 detailed parasitics, so PrimeTime uses the lumped 5 ohms and silently ignores the SPEF resistance for that net. The completeness check passes, timing looks plausible, and nothing errors, yet long_route delay is computed from a stale hand-set value rather than the real extraction, which can either mask a genuine violation or manufacture a fake one. It survives every obvious check because the failure mode is not missing data - it is shadowed data, and -check does not flag shadowing. Catching it requires a deliberate shadowing audit: cross-reference report_annotated_delay and the lumped-override reports against the nets you expect the SPEF to own, or report_net on the suspect net and notice that its resistance does not match the SPEF. That is exactly why verifying precedence shadowing is a distinct signoff gate from verifying completeness.
In Pd / Signoff
At signoff, after reading the SPEF, an engineer diffs the list of nets carrying lumped R or C overrides against the nets the extraction is supposed to own, and runs report_net on suspicious long routes. Any hit means the constraint file is edited and timing rerun, because otherwise the ECOs generated from that run are fixing the wrong net.
Common Trap
Believing a passing report_annotated_parasitics -check means the extracted values are in use. Completeness and precedence are separate properties: the check confirms an RC network exists on the net, not that the tool chose it over a higher-tier lumped override.
Expect Next
How would you script an automated shadowing audit that flags every net where a lumped override outranks available detailed parasitics?
Topic And Primetime Commands
Back Annotation / SPEF & Parasitics. set_resistance, get_nets, report_annotated_parasitics, report_annotated_delay
Keep building interview depth
Continue practising STA
Continue learning free
Get a practical low-power chapter
Receive the existing “Low Power and Multivoltage Fundamentals” PDF chapter and its download link by email.
Continue practising