What is an extraction corner, and why does signoff check RCmax and RCmin separately?
From PDVerse STA Mentor Guide · pdVerse Mentor Guide
Short Answer
An extraction corner describes an assumption about how wide, thick, and closely spaced the manufactured metal wires actually turn out to be, since real fabrication varies within a controlled range around its target geometry. RCmax models the resistance-and-capacitance combination that makes wires act slowest for a given check, and RCmin models the combination that makes them act fastest, and signoff checks both because a wire varying toward one extreme is not guaranteed to vary the same way as one varying toward the other. This is separate from a PVT timing corner, which covers process, voltage, and temperature variation in transistors, not wires.
Technical Explanation
Setup and hold need opposite ends of the wire's own RC range, not just one "worst" number.
- Fabrication of metal layers can produce wires slightly wider or narrower, thicker or thinner, than their drawn target, and extraction models a range of these outcomes as separate RC corners.
- A thinner, taller wire has more resistance but can have less coupling capacitance, so "worst for resistance" and "worst for capacitance" are not always the same outcome.
- Higher resistance generally slows a signal's own delay, the corner relevant to setup, while lower resistance and higher effective capacitance can make a signal switch faster, mattering more for hold and crosstalk.
- Signoff runs a matrix of RC corners and PVT timing corners together, since a chip must work with worst wires and worst transistors, and best wires and best transistors, not just one axis.
- SPEF files are generated per RC corner, so checking three RC corners needs three separate extractions of the same routed design.
Common Mistake
The Trap: assuming a single "worst-case" RC corner covers both setup and hold risk, since both sound like they want the "slowest" wires.
- Setup generally needs the RC corner that makes delay largest, RCmax, while hold generally needs the one that makes delay smallest, RCmin, so applying only one leaves the other check under-covered.
- Skipping RCmin extraction because RCmax "sounds more conservative" is a common way a hold violation reaches silicon undetected.
Follow-up Question & Model Response
How many RC corners does a typical signoff flow actually need to run, and does that number grow with process node?
Candidate Model Response: Most flows run at least RCmax and RCmin as a baseline, plus a typical RC corner as a sanity check between the extremes, so three is a common starting point. As a node shrinks, metal geometry becomes harder to control relative to its own size, so the RCmax-to-RCmin gap widens and some flows add intermediate corners to catch combinations a two-corner sweep would miss. The exact count trades coverage against runtime.
Practical Example
On a design with an RCmax and an RCmin corner, the same 1mm data wire reports 95ps of delay at RCmax and only 55ps at RCmin. A setup check passing comfortably at RCmax can still hide a hold violation that only appears at RCmin, since the faster delay leaves less margin against a requirement that assumed data would arrive later.
Complete STA Handbook
Master Signoff-Ready Static Timing Analysis
Get the complete 10-chapter STA handbook covering setup/hold margins, clock modeling, OCV/POCV, crosstalk noise, and PrimeTime closure.
Continue practising