The dawn of predictive chip yield design: Along and beyond the memory lane

Abstract

A number of researchers described a mixture importance sampling (MixIS) methodology to enable yield-driven design and extends its application beyond memories to peripheral circuits and logic blocks. The sampling methodology was developed by researchers as a fast Monte Carlo technique developed for memory analysis. The MixIS was a comprehensive and computationally efficient method of estimating low failure probabilities of SRAM designs. This method relied on distorting the Monte Carlo sampling function to produce more samples in the important regions and rare-failure-event critical regions. The MixIS methodology was universal and its efficiency was independent of the underlying technology or application. The methodology was also used to compare two different local bit-select circuits in 45-nm technology.