Design of an Ultra-Wide Stopband DMS Microstrip Low-Pass Filter and Optimization With Tab-Transformer Neural Network

It is difficult to balance the requirements of wide stopband, low loss and miniaturization in microstrip low-pass filter design. To solve the problem, a combination of loading the defected metal structure (DMS) units and Tab-Transformer neural network is adopted in this paper. Firstly, the T-shaped DMS unit is introduced into the microstrip metal surface of the high and low impedance line microstrip filter, and the structure optimization of the DMS unit is carried out by using the Tab- Transformer neural network algorithm, which significantly improves the passband performance and reduces the size of the filter. Then the defected ground structure (DGS) unit is loaded on the ground plane of the filter. By fine-tuning the size of the DGS and the microstrip parallel branch, the harmonic of the filter stopband is further suppressed and the pass-band loss of the filter is also reduced. The test results show that the optimized design of the filter has a passband cutoff frequency of 2 GHz, a passband insertion loss of ≤ 0.1 dB, a return loss S11 reduced to -15.49 dB, and a stopband rejection S21 below -20 dB in the frequency range of 2.73 GHz-15.39 GHz, whereas the area of 20 mm* 18.75 mm meets the anti-harmonic interference of a certain L-band radar, low loss and miniaturization requirements.