A novel design and optimization approach of RF MEMS switch for reconfigurable antenna using ANN method

This paper gives the construction and parametric analysis of low loss RF MEMS switch at low DC frequencies up to 8GHz which support reconfigurable antenna devices. The switch used here exhibit excellent RF characteristics in terms of power and current consumption with typical pull-in voltages lies between 1.912 to 2.125V. The switch also exhibit low insertion loss in the range of (0.01 to 0.03dB at DC-3.0GHz and 0.03 to 0.07dB at 3.0-8GHz) in on condition and with reasonable isolation (9dB at DC- 3.0GHz and 9dB to 17dB at 3.0-8.0GHz) at off position. Further the return loss in up and down states are calculated as -40 to -32dB and less than -0.16dB at DC-8GHz respectively. A feed-forward back-propagation (FFBP) algorithm of neural network is also presented for the validation of changing the anchor arm length, width of upper beam and dielectric of RF MEMS switch. The algorithm assembles new data samples during training from finite element method (FEM) based electromagnetic simulation tool HFSS. Results of the ANN are compared with those of the electromagnetic solver. These physical dimensions are varied and design is optimizing for low power consumption and to achieve acceptable level of s-parameters. The developed algorithm allows the optimisation solutions of the design to be carried out by replacing repeated simulations and also provides lesser processing times whilst still retaining an excellent accuracy as compared with finite element modelling. The optimized isolation and insertion loss results are in very good agreement up to 99.5% as compare with the theoretical results.