Quasi-one-step Gauss-Jacobi method for large-scale interconnect analysis via RLCG-MNA formulation
Quasi-one-step Gauss-Jacobi method for large-scale interconnect analysis via RLCG-MNA formulation
复制标题
通过 RLCG-MNA 公式进行大规模互连分析的准一步高斯-雅可比方法
DOI：
10.1109/isqed.2006.109
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
H. Asai
中科院分区：
文献类型：
--
作者：
Y. Tanji;Takayuki Watanabe;H. Kubota;H. Asai
一步或两步高斯-雅可比方法有效地应用于大规模互连分析。为了使高斯-雅可比方法在每个时间步长迭代1次或2次，互连网络采用RLCG-MNA公式。数值算例表明，一步高斯-雅可比法比Berkeley SPICE快1035倍，比最近被称为快速仿真方法的INDUCTWISE快几十倍。进一步从效率和精度的角度证明了两步高斯-雅可比方法是合理的
One or 2-step Gauss-Jacobi method is efficiently incorporated to the large scale interconnect analysis. In order to make the Gauss-Jacobi method within 1 or 2 iterations during each time step, the interconnect network is formulated by the RLCG-MNA formulation. In the numerical example, it is illustrated that the one-step Gauss-Jacobi method is 1,035 times faster than Berkeley SPICE and several tens times faster than INDUCTWISE which is known as the recent fast simulation method. Further, we show that the 2-step Gauss-Jacobi method is rational from efficiency and accuracy points of views