CharTM: An Efficient and Accurate Timing Yield Analysis Method for Memory Characterization based on Generalized Pareto Distribution
TimeTuesday, July 12th6pm - 7pm PDT
LocationLevel 2 Lobby
DescriptionPerformance failure has become a major threat for SRAM circuits. In this paper, we propose a unified memory timing yield analysis method, CharTM, which is based on the Generalized Pareto Distribution(GPD). We design an efficient weighted regression based high dimensional meta model to reduce the simulation cost. Moreover, we apply the strategy of adaptive tail fraction selection to improve the accuracy of high sigma tail distribution. The effectiveness of CharTM is verified on both bit cell circuit and critical path of SRAM. The results show high efficiency and accuracy compared with state-of-out methods and commercial tools.