Neural Network Based GaN HEMT Modelling for Millimeter Wave Power Amplifiers

In this paper, a millimeter wave GaN HEMT model is proposed which applies an artificial neural network (ANN) only to the current source of a compact model to address short channel effects and avoid the overfitting problems in ANN. To create this model, pulsed I-Vs/S-parameters measurement data up to 120 GHz were used. The ANN based model was verified in DC, S-parameters up to 120 GHz and large-signal power performances set at the 71 GHz band. In the verification, the proposed model is the first to demonstrate that ANN-based models can extract the large-signal performance of millimeter-wave GaN HEMTs with reasonable accuracy.