The Wiener Filter Applied to EMI Decomposition

This paper presents a DSP method that allows decomposing conducted electromagnetic interference (EMI) emissions from different sources in large power electronics systems. The estimation process is carried out by a number of Wiener filters (WFs), which represent all different converter states. These are determined by initial commutation event conditions and propagation paths layout. Filters are fed by a semiconductor power switch voltage or current waveforms regarded as sources of perturbation. The EMI emissions are measured on the line impedance stabilization network (LISN) terminals. Optimal filter adaptation is carried out in the frequency domain by measuring and computing input and cross-power signal spectra. Analysis of a parallel quasi-resonant dc-link voltage inverter (PQRDCLI) is outlined to distinguish filters assigned for inverter operation from those for external dc/dc converter interaction. Experimental results are given to illustrate the WF estimation quality. The possibility of detailed decomposition of the LISN-EMI waveforms is depicted in both time and frequency domains. Comparative analysis of frequency responses for PQRDCLI-link voltage changes is given.