Designing high-performance microstrip quad-band bandpass filters (for multi-service communication systems): a novel method based on artificial neural networks

Recently, high-performance multi-channel microstrip filters are widely demanded by modern multi-service communication systems. Designing these filters with both compact size and low loss is a challenge for the researchers. In this paper and for the first time, we have proposed a novel method based on artificial neural network to design and simulate multichannel microstrip bandpass filters. For this purpose, the frequency, physical dimensions, and substrate parameters, i.e., type and thickness, of the BPF are selected as the inputs and the S-parameters, i.e. S-11 and S-21,S- are selected as the outputs of the proposed model. Using an accurate multilayer perceptron neural network trained with back-propagation technique, a high-performance microstrip quad-band bandpass filter (QB-BPF) is designed which has a novel compact structure consisting of meandrous spirals, coupled lines, and patch feeds. The proposed method can be easily used for designing other microstrip devices such as filters, couplers, and diplexers. The designed filter occupies a very small area of 0.0012 lambda(2)(g), which is the smallest size in comparison with previously published works. It operates at 0.7, 2.2, 3.8, and 5.6 GHz for communication systems. The low insertion loss, high return losses, low group delay, and good frequency selectivity are obtained. To verify the design method and simulation results, the introduced filter is fabricated and measured. The results show an agreement between the simulation and measurement.

Automated summary

A novel method based on artificial neural network is proposed for designing and simulating multichannel microstrip bandpass filters. This method utilizes a multilayer perceptron neural network trained with back-propagation technique to design a high-performance microstrip quad-band bandpass filter.