Combining process and statistical variability in the evaluation of the effectiveness of corners in digital circuit parametric yield analysis

Abstract

This paper focuses on two main types of MOSFET variability - systematic (process) and statistical (random) variability and discusses the use of process corners as a measure of yield and circuit performance. We provide a methodology for performing large-scale statistical SPICE simulations as a means of evaluating the accuracy of corners in a system dominated by statistical variability and then expand the methodology to include both systematic and statistical variability within the same large-scale SPICE simulations. This large-scale statistical/systematic approach is compared to the “global + local” statistical corner approach, which consists of statistical simulations around the process corners. Finally 2D kernel density estimates are used to extract yield data from the statistical simulations to allow energy/delay/yield optimization to be performed. This in turn highlights the deficiencies of the statistical corner approach.