论文部分内容阅读
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。
An improved method of static timing analysis is proposed in this paper. The improved quadtree model is decomposed by the random variable of on-chip process variation parameters, and then the exponential function equations of multi-layer spatial distribution are established to obtain the adjacent and next-neighboring patches The fitting coefficients of influence between the two factors make the non-independent random variables into a series of linear additive forms of random variables independent of each other. Finally, the covariance matrix that characterizes the spatial relationship of process parameters in the slice is traversed. Compared with the Monte-Carlo method and the Minnssta method, the accuracy of the improved method is verified, and the effectiveness of the method in reducing the complexity of on-chip non-independent spatial relations is also demonstrated.