Innovative Combline RF/Microwave Filter EM Synthesis and Design Using Neural Networks

This paper presents a novel approach for the design of combline filters using multi stage artificial neural networks (ANN). The method takes advantage of direct synthesis of subsections of the filter using very fast and accurate neural models based on exact electromagnetic (EM) simulations using ANSOFT HFSS. The neural network synthesis approach takes into account the loading effects from adjacent resonators to achieve optimal results. The overall ANN system is capable of designing different filters at different center frequencies and bandwidths. This approach offers significantly reduced time in the design of filters within the validity range of the ANN system. The method has been proven to work efficiently for 4-pole elliptic and 6-pole Chebyshev examples for different center frequencies and bandwidths.