A temperature dependent large-signal drain current neural model for the dual-gate MESFET

In this paper, we present a temperature dependent large-signal drain current neural network model for the dual-gate MESFET. We have modeled an on-wafer symmetric 6×100 μm dual gate MESFET manufactured by Nortel Networks. The measurements of the drain current were taken in a wide range of DC bias points (for V_gs1, V_gs2 and V_ds) and over different values of device temperature. Device temperature is set by mounting the wafer on a temperature controlled thermal chuck. We have tested three neural model types, a three-layer, a four-layer, and a five-layer perceptron neural models. It was found that the five-layer´s model with 21 neurons gives better results for similar number of neurons than those of the three and the four layers. The five-layer model showed an excellent fit to the measurement data. The model error is less than 1%.