Hybrid PKI empirical-neural bias dependent noise model of microwave transistors

In this paper a novel type of bias-dependent hybrid empirical-neural noise model of microwave transistors is proposed. It consists of an empirical noise model based on equivalent circuit representation and a PKI artificial neural network that produces values of the four noise parameters for an arbitrary bias point and frequency on the basis of the noise parameter values obtained by the empirical noise model. As it is shown by the numerical example of pHEMT device, this model provides results that agree well with the device measured characteristics.