Relational SI/PI-Database for a Data-Driven Approach to PCB Design Automation and Performance Prediction

The introduction of machine learning (ML) methods into the design process of printed circuit boards drives the need for large quantities of readily available data. This paper addresses the problems of engineers to find ML-ready data that can be easily reused and combined to enhance printed circuit board (PCB) design by storing the defining parameters in a normalized format within a relational database It implements search and filter functions to obtain relevant data quickly. The database contains data that was used to address a variety of different signal integrity (SI) and power integrity (PI) related problems. Details of the database structure, necessary data conversion steps, currently stored datasets, and a statistical analysis there of are described. This database is capable to be automated to a degree that ML agents can interact with it.