Non-Gaussian Uncertainty Propagation in Statistical Circuit Simulation

Qian Ying Tang and Costas Spanos
EECS, UC Berkeley


Abstract

The mixture of random and systematic variability in state of the art IC technologies often results in non-Gaussian distributions of key performance parameters. A Mixture of Gaussian (MOG) variance propagation scheme is proposed to estimate the transistor-level circuit performance distribution without costly Monte-Carlo simulations. In the proposed method, interval representations and moment matching algorithms are used to find the MOG of the circuit performance.