On the Parameter Identification of Cascaded Behavioral Models for Wideband Digital Predistortion Linearization

This paper discusses the best identification approach to estimate the parameters of cascaded (CC) behavioral models for digital predistortion (DPD) linearization of high efficient wideband power amplifiers. The commonly used least squares (LS) identification method is compared to the proposed gradient descent (GD) -based optimization approach. Experimental results considering a 5G new radio (NR) test signal with 100 MHz instantaneous bandwidth will prove the benefits of training the CC DPD parameters with GD instead of LS, for linearizing a high-efficient pseudo-Doherty load modulated balanced amplifier operated at 2 GHz RF frequency delivering 40 dBm mean output power with around 50% power efficiency.