Broadband PA design using Bayesian algorithms with different covariance functions

Abstract

AbstractThis article examines the effects of Bayesian optimization (BO) with different covariance functions on the optimization of a radio frequency power amplifier (RFPA). An initial PA is designed based on a Chebyshev low‐pass topology using a 10 W Gallium Nitride (GaN) transistor. The objective function is established in a novel manner. The performance of the initially designed PA was optimized by using various BO algorithms, including two different acquisition functions and five different covariance functions. Both simulation and measurement results indicate that the kernel squared exponential (KSE) and Matérn32 covariance functions provide the best option for optimizing PA by BO. A broadband power amplifier (PA) operating at a frequency range from 2.5 GHz to 3.5 GHz has been developed with an output power (Pout) greater than 40.8 dBm and a power‐added efficiency (PAE) greater than 65%.