Evaluating the accuracy of SRAM margin simulation through large scale Monte-Carlo simulations with accurate compact models

Statistical variability due to the discreteness and granularity of charge and matter has a large impact on SRAM performance due to its stochastic nature. In this paper we have performed 5 million SRAM dynamic write simulations with accurate compact models which capture all aspects of statistical variability and use them to benchmark the accuracy of Gaussian threshold voltage modeling strategies and a common SRAM margining technique, MPV. The results show that while MPV and Gaussian VT are proven approaches, deep into the tails of the distribution NPM simulation may present significant opportunities for improved design.