Non-linear Microwave Transistor Modeling through ANNs: a Frequency-domain CAD Implementation

This contribution presents a black-box approach based on Artificial Neural Networks (ANNs) for the accurate and computationally efficient non-linear modeling of microwave transistors. The presented model, here applied to a MESFET test-case, adopts a feed-forward ANN to relate the harmonic components of the device's incident and reflected port waves with arbitrary harmonic loading. As such, it is conceived to be exploited for non-linear design within RF CAD tools based on the Harmonic Balance approach. The detailed implementation within Keysight Advance Design System (ADS), resorting to built-in custom modeling features, is also presented along with model validation results, demonstrating the potential of the proposed approach for fully-black-box non-linear device modeling.