Direct Synthesis Approach for Designing High Selectivity Microstrip Distributed Bandpass Filters Combined with Deep Learning

Abstract This paper presents a direct synthesis and design method based on modified Chebyshev function for high-selectivity wideband bandpass filters (BPFs). This method can synthesize high-order filters with sharp selectivity without requiring complex topologies. Out-of-band rejection of the synthesized filters can be controlled by introducing or removing transmission zeros at finite frequencies. A structure parameter estimation method based on the neural network is used to optimize the structures of BPFs, eliminating the tedious work of manual optimization. The designed bandpass lumped circuits are transformed into distributed circuits accurately by being divided into four basic blocks. In order to prove the performance of the proposed function and the validation of the theory, a five-pole BPF and a nine-pole BPF were designed and measured, proving the practicability of the proposed method.