Welcome to our site! EDAboard.com is an international Electronic Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.
According to my knowledge, four corner STA means:
1. Setup Worst - setup timing in slow corner
2. Setup Best - setup timing in fast corner
3. Hold Best - hold timing in fast corner
4. Hold Worst - hold timing in slow corner
you all missed my question. I know the simple definition and from the posts, it seems that everyone else understands this too!.
Now lets push a little deeper. If we do our timing analysis at slow-slow and fast-fast. We are assured that we covered the two extreme corners and therefore have analyzed everything in between these two corners. Namely slow-fast and fast-slow. Since the PVT at slow-slow corner are the highest and in fast-fast are the lowest it ever can be. It is not possible to find a setup problem in slow-fast that slow-slow analysis missed. Now, the question "Why bother with the two middle cases?"
Four corner checking is the combination of slow or fast corner for P channel and slow or fast corner for N channel.. At least that's the definition our circuit design team had long time ago.
There are cases that SF corner(S: P channel, F: N channel) can be more prone to hold viols than FF.
For example, consider the path that is made with the series of transmission gates. Clock tree is made with buffers or inverters, where both of P channels and N channels are equally exercised, so that SF corner gives you larger clock skew than FF corner. Now suppose the data path has many transmission gates connected serially. As you know, P channel of transmission gate doesn't contribute to the delay much since the major role of P channels in the transmission gates is just pulling up the node voltage beyond the threshold of N channel. So, the delay of transmission gates heavily depends on N channels where the fast corner is used for.
In this scenario, you have larger clock skews, but not much change on data path delay, comparing to FF corner. What does it mean ? Apprarently, it means SF corner is more prone to the hold violation than FF.