EFFICIENT ANALYSIS OF CANONICAL PIECEWISE-LINEAR RESISTIVE-NETWORKS

Abstract

A new algorithm for solving a set of canonical piecewise-linear equations with linear partitions is presented. When solving a set of linear equations repeatedly, the proposed algorithm compares images of next-guess candidates with those of the actual solution while minimizing the computational overhead and finds the best next-guess which is closest to the actual solution. Analysis of test circuits shows that the proposed algorithm requires 8 to 20 times fewer iterations and 5 to 10 times less CPU time than the Katzenelson algorithm. Improved results vary depending on the size of the test circuits. The experimental results also show that the efficiency of the proposed algorithm over the Katzenelson algorithm increases as the number of the piecewise-linear regions of the resistive networks is increased.