An Accurate and Efficient Yield Analysis for SRAM dynamic metrics Using Differential Evolution Algorithm

Abstract

Accurately and efficiently estimate the extremely small failure rate of the SRAM cell is a challenging issue. In this paper, we develop an importance sampling with the differential evolution algorithm to extend the minimized norm importance sampling and mixture importance sampling. The key idea is searching the optimal shift vector (OSV) in the complex and nonconvex failure region using the improved differential algorithm efficiently. Then we construct a mixture of an original distribution and uniform distribution and a distort distribution based the searched OSV to estimate the failure rate of SRAM cell accurately. Our experiment results of a 40nm SRAM Cell show that the proposed method has achieved 1.31x-1.44x speedup than state-of-art methods and 7196x than Monte Carlo (MC) method with less than 2% error compared with MC result when the high accuracy is required (e.g., the relative error defined by the 99% confidence interval reaches 5%).