A Novel Hybrid Algorithm for Source Reconstruction Method in Near-field
Prediction
Abstract
Advanced packaging in electronic systems presents new challenges for
electromagnetic interference issues. The source reconstruction method
(SRM) based on near-field scanning provides a solution for locating
electromagnetic interference sources and reconstructing the
electromagnetic field inside the package. The traditional SRM based on
least squares methods relies on phase information, leading to expensive
measurement facilities and complex testing processes. As a result,
phaseless SRMs with lower testing requirement have become a research
hotspot. However, these methods require solving a nonlinear equation,
which lacks an explicit solution and poses difficulties in extracting
the equivalent radiation source. To address this issue, a new phaseless
SRM that achieves high precision and efficiency is proposed. The method
combines the advantages of differential evolution (DE) algorithm with
the covariance matrix adaptation evolution strategy (CMA-ES) algorithm,
offering fast convergence speed and high accuracy. Compared to
conventional DE algorithm, the proposed hybrid method reduces the error
of the reconstructed field on an average of 9% and improves the
accuracy of the predicted field from 82% to 85% while accelerating
convergence.