Nonlocal Noise Removal Technique to Improve the Electromagnetic Field Pattern in Near-Field Scanning of a Device

A nonlocal noise removal technique, known as block-matching 3-D (BM3D), is introduced to improve the quality of a blurry electromagnetic pattern generated from near field of a device. Noise distribution in electromagnetic pattern is analyzed. Taking the nonlocal self-similarity into account, the BM3D technique removes the noise of the pattern in a 3-D transformation domain based on sparse representation. First, the noise removal technique was validated well on the profile recovery of weak emission of a microstrip line driven by a square wave. Then, the BM3D technique is applied to improve source localization in a field programmable gate array (FPGA) and also to improve pattern clustering in a double data rate (DDR) synchronous dynamic random access memory (SDRAM) integrated circuit (IC). Both the applications demonstrate the effectiveness of the BM3D noise removal technique in the near-field scanning patterns of a device.

I. Introduction

When electromagnetic field is emitted from integrated circuit (IC), the electromagnetic pattern of the near field can be detected using electromagnetic probes. Electromagnetic pattern obtained in near-field scanning over the surface of a device has attracted a lot of attention in the electromagnetic interference diagnosis for EMI source localization [1], EMI diagnosis [2], and immunity test [3]. The mechanism for the generation of the electromagnetic pattern is wave propagation, which is helpful to further understand the properties of waves inside ICs. How to pinpoint what the culprits are that cause such propagation especially for a device with complicated metal interconnect inside it being studied. The source localization method [4] based on the electromagnetic pattern is of general concern for IC design, and pattern clustering [5] could be helpful for the finding of the electromagnetic emission rule.