Design of Microstrip Low-pass and Band-pass Filters using Artificial Neural Networks

Usually the design of microstrip filters is done using simulators and classical approximation methods as Butterworth and Chebyshev, these techniques takes a lot of time to run for designing filters. In this paper we develop a faster artificial neural network model for designing a microstrip low-pass and band-pass filters for all rang of operating frequency , when the input are the dimensions of filter, operating frequency, the features of substrate, and the output are the transmission and reflection coefficients. The database uses for training this model is generated by a linear simulator based on circuits model. Two filters designed by the developed model are a stepped impedance low-pass filter with a cut-off frequency of 1 GHz and a parallel coupled-line band-pass filter with fractional bandwidth of 25 % and a central frequency of 2.45 GHz. The results of simulation are compared with desired results and the effectiveness of this method has been proven.