How does AOCV actually reduce pessimism compared to flat OCV derating, and what's the real cost of using it?

Short Answer

Flat OCV derating applies one derate factor to every cell/net delay in the early direction (hold) and one in the late direction (setup), regardless of path depth or distance -- simple but pessimistic, since it assumes worst-case variation compounds identically everywhere. AOCV (Advanced OCV) instead makes the derate factor a function of logic depth and/or physical distance, since variation statistically partially averages out over more stages/more distance -- less pessimistic, but it requires real AOCV characterization data (depth/distance-vs-derate tables) the library has to actually supply.

Technical Explanation

• Flat OCV derating applies one derate factor to every cell/net delay uniformly, regardless of path depth or physical distance -- simple, but pessimistic, since it assumes worst-case variation compounds identically at every single stage.
• AOCV (Advanced OCV) makes the derate factor a function of logic depth and/or physical distance instead, because variation statistically partially averages out over more stages or more distance rather than compounding fully in the worst direction every time.
• The real cost: AOCV requires actual characterization data -- depth/distance-vs-derate tables the library has to supply -- flat OCV needs no such library data, just two numbers (early/late derate factors).
• Two evaluation modes exist for AOCV: graph-based (fast, uses worst-case-merged path info, still somewhat pessimistic) and path-based (accurate, uses a real recalculated single path, more expensive).
• Flat OCV derating is applied with set_timing_derate -early -cell_delay <value> / set_timing_derate -late -net_delay <value>; the AOCV-specific table-loading command (likely read_ocvm or library-side AOCV tables) should be verified against the ICC2 Implementation User Guide before being asserted as exact syntax.

Follow-up Question & Model Response

"Why would path-based AOCV be more accurate than graph-based AOCV specifically at a reconvergent point in the timing graph?"

Candidate Model Response: Because graph-based analysis merges worst-case delay/slew from different fanin paths at a reconvergent point without knowing the TRUE single-path depth that led there, while path-based analysis recalculates the actual specific path, giving it the real depth/distance to apply AOCV derating against.