Neural network-based tunable microwave filter design for re-configurable biomedical hardware

A knowledge-based method for the design of tunable microwave filters for biomedical applications is described. The method uses a backpropagation neural network (BNN) for mapping RF filter responses to filter resistor, capacitor and inductor values. The BNN acts as an efficient circuit reconfiguration tool which when supplied with a set of filter responses calculates the required component values with high accuracy. To demonstrate the efficacy of the approach, a microwave high pass filter was tuned from 7GHz to 10GHz in steps of 0.2GHz. Filter hardware is reconfigured during tuning using varactor diodes, MOS resistors and tunable micro-inductors.