Investigation on timing analysis inaccuracies

Timing analysis inaccuracies due to crosstalk, multiple gate input switching, supply voltage variation, temperature, manufacturing variation, etc. are very common. How do you tackle them in real life designs? How do you do it using Prime Time (PT)?

Randomized algorithm/approximation scheme for statistical timing analysis

Statistical timing analysis gives a distribution for signal delay at each node in a netlist. A Monte Carlo simulation can give discrete distribution functions. Can there be a randomized algorithm or approximation scheme for statistical timing analysis with guaranteed error bound?

Process variation extraction

Process variation extraction needs to be design specific, and based on random field simulation, as mechanical engineering techniquesStatistical timing analysis is based on a priori knowledge of process variations. The lack of such a priori knowledge of process variations prevents accurate statistical timing analysis and has been largely blamed for foundry confidentiality policy. In this paper, it is shown that a significant part of process variations are specific to the design, and can only be achieved based on production chip performance variabilities.