Design of Ku-Band Multiplexer with High Power Dielectric Resonator Filters Using Neural Networks

In this paper, the recently developed modeling technique for high power dielectric resonator (DR) filters based on neural network modeling of simplified generalized scattering matrix (GSM) of filter junctions is applied to the design of a Ku-band multiplexer with DR channel filters. As demonstrated, the modeling technique is highly efficient for multiplexer design, and requires significantly less time than conventional EM-based method. The reflection coefficient and transmission coefficient of the model show a good match to EM simulation results in both near-band and wide-band frequency ranges. The physical dimensions obtained from the developed model also have comparable accuracy to EM results.