Python-LTSpice Framework for Multi-Objective EMC Filter Optimization

EMC filters should ensure that a device complies with respective regulations, but at the same time should be as small and cheap as possible. Besides the compliance with EMI limits, all used filter components also need to withstand the rated current and voltage of the specific application. For real products this constitutes a discrete valued optimization task, as components and especially common-mode chokes (CMC) are usually not available in all sizes. This work proposes the application of a sorting genetic algorithm (NSGA-II) for described optimization problem. The presented workflow builds on freely available tools, namely LTSpice for circuit simulation and the Python libraries PyLTSpice and pymoo. For demonstration, a commercially available EMC filter design kit is used. From 696 possible component combinations the optimal common-mode (CM) filter for a frequency range of 9 kHz - 100 MHz is found, considering price, volume, ampacity and PCB parasitics.