Noise Source Localization with Sparse Magnitude-Only Magnetic Field Measurements Using KNN Algorithm
Journal
Final Program - 2026 Asia-Pacific International Symposium and Exhibition on Electromagnetic Compatibility, APEMC 2026
Start Page
1
End Page
4
ISBN
[9798331566593]
Date Issued
2026-05-04
Author(s)
Abstract
This study proposes a method for locating an electromagnetic interference (EMI) source based on the KNearest Neighbors (KNN) algorithm. This method can achieve fast and accurate dipole localization using a set of sparse nearfield magnetic field measurement data (without phase information). Compared with conventional high-density scanning approaches, it significantly reduces measurement complexity and time consumption. Through comprehensive simulation and experimental analyses, three key factors affecting prediction performance were identified: (1) probe height; (2) spacing between measurement points; and (3) parameter settings of the KNN training dataset. For integrated circuits (ICs) with a sensing area around 13.2 mm × 13.2 mm, given 3 × 3 probing points with a grid pitch of 4.4 mm, the optimal configuration determined in this work is a probe height of 2 mm. Simulation results show that even in a noisy environment with a signal-to-noise ratio (SNR) of 10 dB, the method achieves a localization accuracy of over 70% within an error range of 1 mm. Experimental results show a consistent trend with the simulation results, validating the robustness and practical feasibility of the proposed method.
Event(s)
2026 Asia-Pacific International Symposium and Exhibition on Electromagnetic Compatibility, APEMC 2026
Subjects
Electromagnetic interference (EMI)
equivalent dipole model
K-nearest neighbors (KNN) method
near-field scanning
radiated emission sources
source reconstruction
Publisher
IEEE
Type
conference paper