Extraction of equivalent sources from near-field scanning data with a restart differential evolution method

Abstract

Complex electronic circuitry is seen in many devices, which results in a vast interference potential. To predict the electromagnetic interference, source reconstruction techniques that extract equivalent sources of current or dipole from scanned near-field (NF) data are developed. Some source reconstruction techniques need the knowledge of NF phase and they employ the least-square method to solve inverse problems (Z. Yu, et al., IEEE Trans. EMC, 1, 2013, 97–108). Due to the ill-posted nature of the problem, regularization techniques are often needed along with the least-square method to eliminate potential false solutions. However, accurate measurement of NF phase is usually difficult in practical applications, especially as the operating frequency is very high. Some other source reconstruction techniques which do not need the NF phase are also developed. They employ global optimization algorithms such as the genetic algorithm to solve inverse problems (X. Tong, et al., IEEE Trans. EMC, 2, 2010, 462–470). These techniques are generally associated with high computational complexity, and they need several hours to obtain the equivalent dipoles of a simple printed circuit board (PCB). An effective source reconstruction technique is proposed in this abstract for the prediction of radio-frequency interference (RFI). The proposed technique does not need to know the NF phase. Equivalent magnetic dipoles are used to model the radiated emission sources of devices, and they are determined by solving an inverse problem with a proposed restart differential evolution (DE) algorithm. Compared to previous methods, the present method has much lower computational complexity. It only needs around 10 minutes on normal desktop PC to extract the equivalent dipole sources of a PCB.