Ultrawideband Schiffman Phase Shifter Designed With Deep Neural Networks

This article presents a novel method for the forward modeling and inverse design of class Schiffman phase shifters using deep neural networks (DNNs). Since DNNs are capable mapping highly nonlinear correlations between inputs outputs, we constructed fully connected DNN to predict electromagnetic (EM) responses given their physical dimensions. Based on this fast accurate tool, cascaded was then built trained achieve instant on-demand shifter designs. approach is versatile can be easily modified accomplish different goals. To demonstrate its efficacy, two realize designs with bandwidths (40% 60%) arbitrary shift targets (0°–180°). Simulation results experimental verifications substantiate that performances comparable state-of-the-art Moreover, discussed proposed methods’ potential in dealing tasks nonintuitive beyond scope existing approaches. We envision extended various EM components including but not limited antennas, filters, power dividers, frequency selective surfaces (FSS).