Neural Modeling of High-Frequency Forward Transmission Coefficient for HEMT and FinFET Technologies

This paper is devoted to examining the ability of artificial neural networks to model the forward transmission coefficient, which represents an important figure of merit for microwave transistors. This analysis is carried out for two different on-wafer devices, namely GaAs HEMT and Si FinFET. As far as the HEMT technology is concerned, the model is developed for three devices which differ in gate width. For the FinFET technology, the model is determined not only for the whole device but also for the actual transistor by using the de-embedding procedure to subtract the effects of pads, transmission lines, and substrate from the measurements. The obtained models have been developed and validated in a wide range of bias conditions for a frequency range up to 50 GHz.