GaN HEMT small-signal modelling: Neural networks versus equivalent circuit

This paper presents on overview on the microwave small-signal modelling for the GaN HEMT technology. The two approaches contrasted here are: the standard modelling approach based on the equivalent-circuit representation and the behavioural modelling approach based on artificial neural networks. These two approaches are applied to a GaN HEMT device working in a wide range of biases and at different ambient temperatures, ranging from 20°C to 80°C. The pros and cons of both approaches are examined in terms of accuracy, complexity, and usefulness of the achieved models.