A Machine Learning-Based Microwave Device Model for Fully Printed VO2 RF Switches

Fully printed vanadium dioxide (VO2) based Radio Frequency (RF) switches have been recently developed for advanced frequency-reconfigurable RF electronics. A reliable and versatile model for the VO2 switches is required for design and simulations in the modern Computer-Aided Design (CAD) tools. This paper proposes a machine learning (ML) based model for VO2 RF switches, which is much more time and resource efficient as compared to the traditional device models. The computational efficiency, accuracy and robustness of the proposed model over a frequency range of 30 GHz is demonstrated through an excellent agreement between the modelled and measured results. The comparison between the measured and modelled results demonstrate a mean-square error (MSE) of lower than 5 x 10-4 and 5 x10-3 for the magnitude and phase values over the complete frequency range.