Design of a novel wideband microstrip diplexer using artificial neural network

In this paper, we use an artificial neural network (ANN) to design a compact microstrip diplexer with wide fractional bandwidths (FBW) for wideband applications. For this purpose, a multilayer perceptron neural network model trained with the back-propagation algorithm is used. First, a novel resonator consists of coupled lines loaded by similar patch cells is proposed. Then, using the proposed ANN model, two mathematical equations for S11 and S21 are obtained to achieve the best configuration of the proposed bandpass filters and tune their resonant frequencies. Finally, using the obtained bandpass filters, a high-performance microstrip diplexer is created. The first channel of the diplexer is from 1.47 GHz up to 1.74 GHz with a wide FBW of 16.8%. The second channel is expanded from 2 to 2.23 GHz with a fractional bandwidth of 11%. In comparison with the previous designs, our diplexer has the most compact size. Moreover, the insertion losses at both channels are improved so that they are 0.1 dB and 0.16 dB at the lower and upper channels, respectively. Both channels are flat with a maximum group delay of 2.6 ns, which makes it suitable for high data rate communication links. To validate the designing method and simulation results, the presented diplexer is fabricated and measured.