ANNs in Bias Dependant Small-Signal and Noise Modeling of Microwave FETs

In this paper an efficient procedure for determination of small-signal and noise behavior of microwave transistors for various bias conditions is proposed. An empirical transistor noise model based on an equivalent circuit (improvement of Pospieszalski’s noise model) is considered. Since it is necessary to extract values of the model equivalent circuit for each bias point (which requires the measured data acquiring and repeated time consuming extraction procedures), it is proposed to use an artificial neural network to model the bias dependence of the equivalent circuit parameters. In that way, 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, the device small-signal scattering and noise parameters are easily obtained for an arbitrary bias point from the device operating range without changes in the model. The proposed modeling approach is exemplified by modeling of a specific MESFET device in a packaged form.