Conducted Susceptibility Visualization and Modeling Optimization for Analog and Mixed-Signal Chips Based on a Cluster Heatmap Approach

Abstract

An automated, sensitive-feature identification–visualization method is proposed in this article to address the challenges—namely, the limited identification accuracy, reliance on single-dimensional static criteria, and bottlenecks in Black box modeling efficiency—in evaluating the conducted susceptibility of analog and mixed-signal (AMS) integrated circuits (ICs) in complex electromagnetic environments. The proposed method is based on the data selection initialization of an improved k-means clustering approach, which does not require predefined output failure criteria for evaluation. The developed method addresses issues inherent to traditional clustering techniques, including sensitivity to initial conditions and random clustering outcomes, by leveraging the chip’s baseline response characteristics under undisturbed conditions as a physical constraint and integrating a multi-indicator fusion algorithm. The proposed method generates heatmaps of electromagnetic interference susceptibility distributions, providing intuitive quantification and visualization of the chip’s susceptibility evolution patterns in the 2-D frequency–power interference space. Experimental results demonstrate that this approach had good consistency with the susceptibility boundaries determined according to traditional standards and, more critically, revealed the continuous degradation trajectory of AMS ICs from the linear operating region through the nonlinear distortion transition zone to the functional failure zone. Furthermore, the partition modeling strategy guided by clustering results reduced data acquisition and model training costs by eliminating redundant test points and focusing on critically sensitive regions while maintaining the black-box model’s prediction accuracy (the model iteration convergence speed improved by approximately 33.7%, and the total time was reduced by 25.9%). This work provides an objective evaluation tool for the electromagnetic susceptibility of AMS ICs without predefined output failure criteria, characterizes the susceptibility response evolution under broadband interference, and establishes a robust theoretical and methodological foundation for chip screening, in situ replacement, and efficient electromagnetic behavior modeling in high-reliability scenarios.