An automatic multi-level solver switching strategy for PEEC-based EMC simulation

Abstract

In this paper, we propose a new, efficient and robust automatic solver switching strategy, called α-SAMG, tailored to linear systems of equations arising in PEEC-based EMC simulation. The PEEC method (partial element equivalent circuit) is an approach for transforming conducting objects into linear networks with basic electrical elements. Such equivalent circuit models can be simulated by conventional circuit solvers such as SPICE (simulation program for integrated circuit emphasis) based on the MNA (modified nodal analysis) approach. By applying appropriate sparsification techniques, sparse PEEC matrices can be obtained, adequate for iterative solvers. By using multi-level approaches, linear complexity w.r.t. time and memory consumptions can be achieved in the best case. Due to the fact that properties of PEEC matrices can differ drastically, we developed the automatic solver switching strategy α-SAMG. Its efficiency for PEEC matrices is demonstrated for seven practically relevant benchmark cases by comparison against standard solvers. In particular, it is shown that α-SAMG provides a robust and fast solution strategy.