基于OMP的SRAM成品率分析的分组建模方法  

OMP-based SRAM performance modeling approach via clustering the simulation data

作　　者：梁堃[1,2] 张龙 赵振国[3] 杨帆[4] LIANG Kun;ZHANG Long;ZHAO Zhenguo;YANG Fan(Microsystem&Terahertz Research Center,China Academy of Engineering Physics,Chengdu Sichuan 610200,China;Institute of Electronic Engineering,China Academy of Engineering Physics,Mianyang Sichuan 621999,China;Software Center for High Performance Numerical Simulation,China Academy of Engineering Physics,Beijing 100088,China;State Key Laboratory of ASIC&Systems,Fudan University,Shanghai 201203,China)

机构地区：[1]中国工程物理研究院微系统与太赫兹研究中心,四川成都610200 [2]中国工程物理研究院电子工程研究所,四川绵阳621999 [3]中国工程物理研究院高性能数值模拟软件中心,北京100088 [4]复旦大学专用集成电路与系统国家重点实验室,上海201203

出　　处：《太赫兹科学与电子信息学报》2020年第1期155-159,共5页Journal of Terahertz Science and Electronic Information Technology

摘　　要：静态随机存取存储器(SRAM)电路的失效是极小概率事件,并且不同电路条件下的失效区域边界可能相距很远。因此,为了更高效地获得更精准的SRAM成品率预测结果,提出一种基于正交匹配追踪(OMP)算法的SRAM性能分组建模方法,并应用于典型SRAM电路成品率的预测。此方法主要根据不同SRAM电路条件下失效区域边界距离的差异将仿真数据划分为多组,之后利用OMP算法对不同组的数据分别建立多项式模型,该模型可用于对SRAM电路的成品率进行快速分析预测。与标准蒙特卡洛统计算法及基于OMP的单一建模方法相比,基于OMP的分组建模方法不仅可以缩短建模时间,提高建模准确度,还能够获得更加精确的SRAM成品率预测结果。The failure of Static Random Access Memory(SRAM)circuits is an extremely rare event,and the boundaries of failure regions under different conditions may differ a lot.Therefore,in order to obtain more accurate SRAM yield prediction results more efficiently,an SRAM performance modeling method in multiple groups based on Orthogonal Matching Pursuit(OMP)algorithm is proposed and applied to the yield prediction of typical SRAM circuits.This method divides the simulation data into several groups according to the difference of the boundary distance of the failure regions under different conditions,and then uses the OMP algorithm to establish the polynomial models for different groups of data,which can be utilized to analyze and predict the yield of SRAM circuit quickly.Compared with the standard Monte Carlo algorithm and the single-model method,the proposed method requires much less time to build models with smaller fitting error,and can achieve more accurate SRAM yield prediction results.

关 键 词：分组建模 正交匹配追踪 重要性采样 静态随机存取存储器成品率预测 

分 类 号：TN492[电子电信—微电子学与固体电子学]