A Modified Decomposed Vector Rotation-Based Behavioral Model With Efficient Hardware Implementation for Digital Predistortion of RF Power Amplifiers

Abstract

This paper proposes a novel hardware implementation strategy to achieve low-cost design for digital predistortion (DPD) of radio frequency (RF) power amplifiers (PAs) using a modified decomposed vector rotation (DVR)-based behavioral model. To make the model hardware friendly, we first modify the model into a sub-decomposed format which significantly reduces the computational complexity in model extraction. We then reassemble the coefficients and propose a simple digital implementation structure for real-time signal processing in the transmit path. A new Dual-Direction COordinate Rotation DIgital Computer (DD-CORDIC) design is also proposed to simultaneously calculate both magnitude and ejθn values to facilitate the model implementation. To validate hardware implementation, a wide-band signal is employed to evaluate the performance with a Doherty power amplifier. Experimental results show that the proposed approach can achieve comparable performance with much lower system complexity compared to that using the conventional approaches.