A new frequency domain waveform relaxation algorithm for PEEC models

Abstract

Waveform Relaxation (WR) is conventionally a time domain algorithm for the solution of linear and nonlinear partial and ordinary differential equation problems. It is especially suited for large systems and parallel processing due to the large compute-To-communication ratio and its ability to break up large problems into smaller ones. In this paper, we investigate the new application of WR to linear frequency domain problems. We show that we can use the insights gained from the extensive time domain WR research results to come up with a new frequency domain approach. For this purpose, we compute frequency response waveforms rather than time domain waveforms. Fortunately, some techniques such as partitioning are similar for both approaches. In this paper, we test some of the new algorithms for the frequency domain analysis. The new approach is aimed at the solution of large problems using single and parallel processor solutions.