Statistical Transient Analysis
Shih-Hung Weng1, He (Vincent) Peng2 and Ernest S. Kuh
The goal of this research project is to develop statistical transient analysis algorithms and tools using parallel processing. With the technology scaling, the parameter variations become more pronounced. The phenomena have a significant impact on the yield, system performance, and power dissipation. Traditionally, the effects of variations are simulated via Monte Carlo method, which tends to be time consuming. We propose to explore incremental and adjoint network methods to derive the envelop of the transient responses with statistical distribution. Given parameter variation models, we calculate the worse cases and the probabilistic of the events according to selected metrics. Parallel algorithms will be devised for matrix solver and differential system integration.
1UC San Diego