Identification of conducted EMI noise based on independent component analysis and signal statistics applied in underground power electronic system

Abstract The conventional compliance test for conductive EMI noise is based on LISN which can only show the pass or failure of the total noise, but can not indicate which component (CM or DM component) is the predominant one if the test fails to meet the EMC regulation, and thus, EMI filter design is difficult due to the lack of CM or DM noise information. Though some attempts have been done for conductive EMI noise separating network research and CM or DM noise can be extracted, the network performance usually is not ideal at high frequency because of the parasitic effects and also the network is often complicated and not cheap in implementation. This paper presents a novel identification method of conducted EMI noise using independent component analysis (ICA) and signal statistics for underground power electronic systems. This Research shows that with this signal process technique, the identification of EMI noise is not only robust, but also simple and economic in implementation.