基于改进四叉树模型的统计静态时序分析  

Statistical Static Timing Analysis Based on Modified Quad-Tree Model

作　　者：潘旻[1] 喻明艳[1] 

机构地区：[1]哈尔滨工业大学国际微电子中心,威海264209

出　　处：《微处理机》2014年第5期1-4,共4页Microprocessors

摘　　要：提出一种改进的静态时序分析方法,该方法通过对片内工艺变化参数随机变量进行改进四叉树模型分解,然后建立多层分布空间关系指数函数方程组求得片内相邻、次邻块间影响的拟合权重系数,使得非独立的随机变量转化为一系列相互独立的随机变量线性相加的形式,最后遍历获取表征片内工艺参数变化空间关系的协方差矩阵。通过和Monte-Carlo方法以及Minnssta方法仿真结果对比,验证了改进方法的精确性,同时也表明了该方法在降低片内非独立空间关系复杂性方面的有效性。In this paper,a modified approach for static timing analysis is presented. It decomposes intra-die random variables based on modified quad-tree distribution model and makes the dependent random variables as a linear sum of independent random variables,by solving the multi-level distributed spatial correlation equations,which are related to exponential functions,to obtain the fitting weight coefficients of adjacent and diagonal intra-die squares. Consequently,the covariance matrix,which represents for the spatial correlations of intra-die process variations,can be derived through traversal. The simulation results,from Monte-Carlo method and Minnssta method,confirm the accuracy of modified approach and show that it is effective in reducing the complexity of analyzing the dependent spatial correlations.

关 键 词：工艺参数变化 空间关系 统计 四叉村 静态时序分析 

分 类 号：TN402[电子电信—微电子学与固体电子学]