Fast and Data-Efficient Signal Integrity Analysis Method Based on Generative Query Scheme

With the rapid development of electronic technology, signal integrity (SI) analysis of high-speed circuit channels has become a challenging task. As SI designs become increasingly complex, traditional SI analysis methods require significant time consumption. To address this issue, we propose a fast and data-efficient SI analysis method called generative query scheme (GQS). In GQS, we initially trained a machine learning-based SI analysis model to predict SI metrics. Then, we design a Markov decision process (MDP) to sample the most informative SI design samples from the design space. The sampled SI simulation data is utilized to continuously train the machine learning-based SI analysis model to improve its prediction accuracy. Training the model via GQS is significantly faster than traditional SI analysis methods, and the prediction accuracy of our model exhibits an obvious advantage compared to the existing SI analysis approaches. While complementing traditional SI analysis methods for precise validation, GQS greatly improves the time of SI design.