A nonlinear MESFET model for intermodulation analysis using a generalized radial basis function network

摘要

In this paper we use a generalized radial basis function (GRBF) network to model the intermodulation properties of microwave GaAs MESFET transistors under dynamic opera- tion. The proposed model receives as input the bias voltages of the transistor and provides as output the derivatives of the drain-to-source current. which are responsible for the inter- modulation properties. The GRBF network is a generalization of the RBF network, which allows different variances for each dimension of the input space. This modification allows to take advantage of the soft nonlinear dependence of the output derivatives with the drain-to- source bias voltage. The learning algorithm chooses the GRBF centers one by one in order to minimize the output error. After selecting each new center from the training set, the centers and vanances of the global network are optimized by applying gradient descent techniques. Finally, the amplitudes are obtained by solving a least-squares problem. The effectiveness of the proposed GRBF model is validated through load-pull intermodulation prediction based on the experimental nonlinear characterization of an NE72084 MESFET device.