Pattern clustering method of magnetic near‐field radiation emissions based on DBSCAN algorithm

Abstract

Abstract The analysis of magnetic near‐field radiation emissions (MNRE) has recently raised more attention in device‐level electromagnetic compatibility testing. Pattern clustering of MNRE for integrated circuits manually is very time‐consuming because of the multi‐dimensional characteristics of MNRE, such as frequency, spatial position, emission intensity, etc. This paper proposes a novel pattern clustering method of MNRE, including strong emission frequency extraction, feature extraction, and density‐based clustering. Ring oscillator and five working states are designed on a Field Programmable Gate Array with 256 Ball Grid Array package, which are used to create a complex multi‐source emission case for verifying the effectiveness of the clustering method. The verification results show that the proposed method can correctly cluster the multi‐source emission patterns. Further, the method is also applied to a Microcontroller Unit with unknown operating states; the results show that the proposed method also can effectively distinguish the unknown emission patterns and locate the unknown interference source. The accuracy of the interference source location is proven by 3D X‐ray microscope inspection.

Keywords

• artificial intelligence
• magnetic field measurement
• measurement systems