A hybrid modeling approach for drain-source capacitance of source field plated GaN FET devices for multi-bias application

The nonlinear drain-source capacitance (Cds) of source field plated power RF GaN device has been modeled by an artificial neural network (ANN) for easy integration with a compact model. This hybrid modeling approach can be suitable for a quick, technology independent solution to a complex problem. In this study it is shown that the proposed hybrid model based on a modified EEHEMT model demonstrates improved large signal accuracy over wide drain bias ranging from 28V to 65V when compared to the fixed Cds used in the stand-alone compact model.