模型降阶算法在互连线系统仿真中的应用

Application of model reduction algorithm for the simulation of interconnect system

在进行集成电路系统的仿真时,如何加快含有互连线寄生效应所产生的延时信息的计算变得尤为重要。采用模型降阶的方式对具有互连线寄生效应的电路系统系数矩阵进行降阶,以达到加快含有互连线延时信息的互连电路仿真速度的目的。通过泰勒级数展开的高阶逼近技术,将传递函数中的e-sτ项进行多项式展开逼近,而后采用高阶Arnoldi算法进行降阶,所以降阶算法继承了传统矩匹配算法的保持无源性和结构性的优点,又能保证一定的精确度。算法最初的目标降阶数采用Hankel奇异值决定,减少了降阶的迭代次数,大大缩减了计算时间。

During integrated circuit system＇s simulation,it becomes particularly important about how to speed up the calculation of interconnection＇s delay. For accelerating simulation＇s speed of interconnect system with delay information,model reduction method is adopted to achieve reduction of interconnected circuit system＇s coefficient matrix. With high-order approximation technique of Taylor series expansion,the transfer function＇s e- sτitem is converted a polynomials,then uses the high-order Arnoldi algorithm to finish reduction. So reduction method proposed inherits the advantages of passive and structural of tradition moment matching algorithm,and has a high precision. Since the algorithm uses the initial optimal target number decided from singular value of Hankel matrix,so the calculation time is decreased by reducing iteration＇s time.