Knowledge-Based Multi-Path Neural Networks for Modeling Through Packaging Glass Vias

This work introduces a novel approach for analyzing signal integrity in coaxial through-glass vias (TGVs) using advanced machine learning techniques. The study utilizes a knowledge-based multi-path neural network to assess the signal integrity of coaxial-TGVs by predicting parasitics like resistance and inductance across varied frequency and temperature ranges. By leveraging a driver-via-load model, the proposed architecture provides accurate assessments of coaxial-TGVs, making it highly suitable for high-frequency, temperature-sensitive applications.