Digital Predistortion of RF Power Amplifier Using the NARX-SVR Model

Abstract

This letter presents the digital predistortion for power amplifiers (PAs) using a new behavioral model that combines a nonlinear autoregressive exogenous architecture with a support vector regression (NARX-SVR) algorithm. Unlike other models based on SVR, the NARX-SVR requires lower support vector numbers (SVN) and floating-point operations (FLOPs) without sacrificing the accuracy of the dynamic nonlinear behavior model of the PA. The NARX-SVR model was validated by modeling and predistorting a commercial GaN hybrid Doherty PA (HDPA) driven with a 2- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times 20$ </tex-math></inline-formula> -MHz LTE signal with an output power of 42 dBm. Experimental results show excellent agreement between measured and modeled AM–AM and AM–PM data. The distortion correction is better than 14 dB, and the error vector magnitude (EVM) is less than 2%. Besides, the FLOPs and SVN are reduced by 18% and 20%, respectively, compared with the dynamic time-delay SVR model.