Support vector regression for harmonic optimization in continuous class-F power amplifier design

We propose an automatic optimization method to control the harmonic impedances of continuous Class-F power amplifiers (PAs) using support vector regression (SVR). The matching networks are optimized to align with the desired harmonic impedances at the package plane of Cree's CGH40010. Two design cases with different desired harmonic impedance trajectories are used to validate the optimization robustness of SVR. Both fabricated power amplifiers achieve efficiency higher than 60.7 % from 1.5 to 2.5 GHz, as well as output power larger than 39.8 dBm with gain better than 12.9 dB. Comparing PA designs reveal that SVR greatly surpasses Advanced Design System's (ADS) genetic optimizer and the simplified real frequency technique (SRFT) while displaying similar performance to Gaussian processes regression (GPR).