Convolutional-Neural-Network-Based EBG Structural Design for High-Speed PCB Noise Suppression

Abstract

With rapid advances in technology, the operating frequencies of digital systems have increased to several GHz bands. This has led to an increase in simultaneous switching noise(SSN). To reduce SSN, electromagnetic bandgap(EBG) structures have been intensively studied. One of the critical steps in the design of an EBG structure is to predict the stopband that reduces SSN. Existing methods include using a 3D electromagnetic field simulation program or equations based on the Floquet theory. However, these have limitations. In this study, we verified a new method for predicting the stopband using a convolutional neural network(CNN). Specifically, a CNN architectural model was used to compare structures that perform well in predicting the stopband. It was also used to confirm that the DenseNet showed high performance.