Artificial neural networks in bias dependant noise modeling of MESFETs

An efficient procedure for accurate noise parameter prediction of microwave MESFETs / HEMTs for various bias conditions is proposed in this paper. It is based on an improved Pospieszalski's noise model. The bias dependences of the noise model elements are modeled by an artificial neural network. Therefore, it is necessary to acquire the measured data and extract the equivalent circuit parameters only for several operating biases used for the network training. Once the neural network is trained and assigned to the considered noise model, the device noise parameters are easily obtained for each bias from the device operating range. It is done without changes in the model and without time consuming and complex measurements and optimizations.