Neural network modeling of active devices for use in MMIC design

This work can be classified into three parts: The first part is a multidime nsional signal-noise neural network model for a microwave small-signal tran sistor. Here the device is modeled by a black box, whose small signal and n oise parameters are evaluated through a neural network, based upon the fitt ing of both these parameters for multiple bias and configuration with their target values. The second part is the computer simulation of the possible performance (F, V-i, G(max)) triplets. In the final part, which is the comb ination of the first two parts, the performance curves are obtained using t he relationships among operation conditions f, V-CE, and I-CE; the noise fi gure, input VSWR and maximum stable transducer gain. (C) 1999 Elsevier Scie nce Ltd. All rights reserved.