AI-Enhanced Modal Decomposition Method for Fast and Efficient PCB Modeling and Signal Integrity

• Published in: 2023 IEEE 7th International Symposium on Electromagnetic Compatibility (ISEMC) pp. 1 - 4
• Main Authors: Gao, Zhaodong, Su, Ming, Guo, Xingyue, Chen, Xilian, Yan, Chuming, Liu, Yuanan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20.10.2023
• Subjects: Artificial Intelligence; Computational modeling; Computer architecture; Data models; Electric potential; Electromagnetic compatibility; Machine learning; Measurement uncertainty; Modal Decomposition; PCB Modeling; Predictive models; Signal Integrity
• DOI: 10.1109/ISEMC58300.2023.10370118

An AI-enhanced modal decomposition method is proposed in this paper for fast and efficient high-speed multilayered PCB modeling and integrity. Depending on modal patterns of PCB local structures, modal decomposition method divides a given high-speed PCB interconnections into independent modal cells, and take advantages of different evaluation methods (analytical, numerical, AI-based) to analyze each modal cell. With the assistance of AI technology, compound methods find a way to compute complex PCB structures effectively in this modal decomposition method. We briefly introduced the architecture and workflow of the proposed method, and then gave a practical application example to show the validity. Data show frequency error of 2.3/1.4% and amplitude error of 0.3/0.6dB for 0-28/28-40GHz between measurement and model prediction, showing great potentials of the AI-enhanced modal decomposition method for PCB modeling and signal integrity.