Signal Integrity Analysis Based On Machine Learning

In recent years,with the rapid development of intelligent manufacturing and chip manufacturing technology and the popularization of 5G technology,electronic equipment is highly integrated,and components and PCBs tend to be miniaturized and refined.The traditional method of predicting printed circuit board(PCB)signal integrity problems has the characteristics of high cost,low efficiency,and unintelligence.With the precision of chip technology and PCB design process,the existing simulation methods are about to encounter the bottleneck in computing performance.Using artificial intelligence methods to reasonably predict and analyze the signal integrity problems of PCB boards will become the future development trend.The thesis mainly conducts in-depth research on the intelligent prediction of PCB board signal integrity.Based on high-speed circuit and signal integrity theory,combined with the study and research of artificial intelligence,an intelligent prediction system for PCB board signal integrity problems is realized.The main tasks to be completed include:1.For the first time,a solution for the intelligent prediction of PCB board signal integrity problems is proposed,which provides a technical framework for the future intelligent prediction and analysis of PCB board electromagnetic compatibility;2.The method of reading files is used to extract the PCB model,using PCB board design source file and the IBIS model file to split the PCB board into three types of active device pins,passive device pins and transmission line segment pins,and use the DFS algorithm to construct a "PCB board sequence model" to quantify all PCB boards,and control the modeling time to the second level;3.For the first time,the machine learning model is used to predict the crosstalk and reflection of the network on the PCB board.The crosstalk prediction has reached an accuracy of 73.2%,and the reflection prediction's error is about 8.5%.By comparing with the experimental results of the simulation software,the machine learning method has a great advantage in performance,reducing the prediction time of signal integrity problems by more than 100 times.