Deep Neural Network Dataset Collection for Optimal Positioning of a Capacitive Compensated Schiffman Phase Shifter

Abstract

This paper presents an accurate analysis and design of a Schiffman phase shifter with optimally positioned compensation capacitances for improved performances. The approach synthesized the microstrip coupler's even and odd mode characteristic impedance and its coupling effect to enhance the phase shift with minimal phase deviation. This was achieved by positioning artificial capacitances along the microstrip directional coupler. The new design method is valid for all substrate thicknesses and permittivity thereby overcoming the bandwidth limitation presented by the minimum spacing between the coupled lines in a conventional microstrip directional coupler. The simulation result demonstrated a maximum phase deviation of ± 3° and a maximum return loss of -0.613 dB. The source and load reflections were well below - 10 dB. The obtained dataset is highly accurate and can be used to train a deep neural network for improved phase error and return loss of a Schiffman phase shifter.