Wideband Schiffman Phase Shifters Designed with Deep Neural Networks

Abstract

Phase shifters play an important role in beam scanning phased arrays, modulators and communication systems. Ideal phase shifters should provide flat phase shift over wide operating frequency band with low insertion and return loss. Schiffman phase shifters are compact in size, easy to fabricate using Print Circuit Board (PCB) technology, while still provide accurate phase shift in relatively wide bandwidth (usually $\gt 50$%), making them stand out from various wideband phase shifter designs. However, due to the approximations being used during the design and the unquantifiable influences of the chamfered entries, fringing effects and parasite inductances of the narrow links between coupled lines, it's hard to analytically calculate the electromagnetic (EM) response of a Schiffman phase shifter given the dimensions. As a result, time-consuming fine-tuning is still required during the design process. In this paper, a novel DNN approach is introduced for the fast inverse design of wideband Schiffman phase shifters. For the first time, a predicting neural network that is capable of simultaneously modeling the phase shift, insertion loss and return loss of Schiffman phase shifter structures over a relative wide spectrum (133%) has been demonstrated. Based on the highly accurate forward predicting network, a tandem inverse design network was also constructed for the fast inverse designs of Schiffman phase shifter with arbitrary phase shift and bandwidth targets. Different from traditional design approaches, the well-trained inverse design network generates design parameters in milliseconds, with no further EM simulation needed. Several Schiffman phase shifters with 60% and 40% fractional bandwidth were designed, fabricated and tested to verify the efficacy of the proposed approach. This DNN-enabled method validates the feasibility of on-demand wideband phase shifter designs, which can be easily generalized to other EM problems, including but not limited to antenna design, microwave circuit design and EM compatibility problems.