Statistical modeling for circuit simulation

Abstract

Robust, high yield IC design requires statistical simulation, and therefore statistical models. Simple "fast" and "slow" sets of model parameters are not sufficient to predict the manufacturing variations of all measures of circuit performance for arbitrary circuit topologies, device geometries, and biases. This paper describes an accurate and efficient approach to statistical modeling and characterization. The procedure is based on physical process parameters, and explicitly accounts for correlated and uncorrelated variations of statistical parameters. The process is generic, and so is applicable to any type of device, and emphasizes the accuracy of device electrical performance variation modeling, rather than model parameter variation modeling. This provides an accurate and simple way to model and simulate the statistical variation of circuit electrical performances.