A Reliable and Fast ANN Based Behavioral Modeling Approach for GaN HEMT

The paper proposes an accurate, fast and advanced neural network approach to model the small signal behavior of GaN High Electron Mobility Transistor (HEMT). The presented approach makes use of the nonlinear autoregressive series-parallel and parallel architectures to model a 2×200μm device for a broad frequency range of 1GHz - 18GHz. A comparison is drawn between the two architectures based on the training algorithm, accuracy, convergence rate and number of epochs. An excellent agreement is found between the measured S-parameters and the proposed model for the complete broad frequency range. The proposed model can be embedded into computer aided design tool for an accurate and expedited design process of RF/microwave circuits and systems.