PrimeTime Timing Closure

timing closure

Hi,

Can anyone please explain me regarding the differences between timing analysis in PrimeTime using SPEF and SDF?

How is the timing analysis using SDF different from timing analysis using SPEF and which one is better?

Regards,
Shyamala

PrimeTime is a static timing analyzer (STA). Internally it actually consists of 2 parts:
The first part is a "delay calculator". As the name suggests, this engine calculates the delay through a gate or the delay along a wire. In order to calculate these delays, PrimeTime needs an electrical-equivalent model for the physical wire. This is generated by another tool called an RC-extractor (e.g.: Star-RCXT from Synopsys). These "RC parasitics", as the wire model is called, are supplied to PrimeTime in the form of a SPEF (Standard Parasitic Exchange Format) file. The resulting delay arcs produced by the delay calculator can be dumped out of PrimeTime in the form of an SDF (Standard Delay Format) file

The second part of PrimeTime is the actual STA engine that computes the slack and the critical paths between all registers, and checks setup times and hold times.

So, it is quite normal for PrimeTime to use a SPEF because PrimeTime cannot extract parasitics itself. But it is very unusual for PrimeTime to take in an SDF. This would only happen if the user had some other delay calculator that he/she trusted more than PrimeTime's own delay calculator.

To get more accurate results, use SPEF. This will allow the tool compute the delays with high accuracy as the detailed parasitics are provided as SPEF.

You can also read in the SPEF and write out a SDF using primetime. Then use the SDF for timing analysis. But the delays will be little off from the SPEF results. If you are looking at "pico second" range delays, i suggest you to go for SPEF based analysis.

usually , use SPEF for PT to run and SDF for GSIM to run!