Nonlinear transient analysis based on wave digital filters

Abstract

Simulation of large circuit networks is a CPU-time and memory consuming task. Computer clusters and multi-core processors have become common but algorithms must be parallelizable to fully take advantage of these architectures. Therefore, it is of great interest to find efficient parallel algorithms for circuit simulation. In this thesis an algorithm based on a Wave digital Filter (WDF) model of the circuit combined with waveform relaxation is presented for the first time. The new approach is tested with the simulation of a nonlinear transmission line. Several variations of the algorithm and the potential for parallelization are analyzed. Topics relevant for the work in this thesis are presented first followed by transient simulation results using a Jacobi-like algorithm applied to two different WDF models of a nonlinear transmission line. The simulation results are shown to be in agreement with the results obtained using a traditional method. The Waveform Relaxation implementation is presented next followed by a set of simulation results and a discussion. Simulation results indicate that the approach requires more CPU time than traditional methods. Some suggestions for future improvements are given in the last chapter of this thesis.