On the determination of neural network based non-linear constitutive relations for quasi-static GaN FET models

By using a neural network approach that takes into account input/output relationship data along with derivative information in the training process, a fast and straightforward methodology to obtain the quasi-static model of GaN FETs is introduced. This method uses data obtained from pulsed I/V and S-parameter measurements to train three different neural networks which model the drain current, as well as the gate and drain charge functions. The ANN-based model is implemented in Agilent's ADS™ and validated by comparing the results to measured data.