| Abstract: | In this paper, the relative advantages of several widely used MESFET and HEMT models have been compared. The nonlinear behaviours of the Curtice quadratic, Curtice cubic, Statz, Materka, Rodriguez, and Chalmers models were investigated through their current–voltage–temperature characteristics. To better fit such characteristics, neural‐based models of MESFET and HEMT were generated using a Levenberg–Marquardt back‐propagation algorithm. Close agreement was observed between simulated results and experimental data. Copyright © 2007 John Wiley & Sons, Ltd. |
