A Nonlinear Memory Polynomial Model for Digital Predistortion of Power Amplifiers With Strong Nonlinear Distortion

In modern 5 G wireless communication scenarios, the power amplifier (PA) is usually driven into the strong nonlinear region, resulting in the conventional full-band (FB) digital predistortion (DPD) couldn't work properly. In this paper, an off-line modified iterative learning control (ILC) based training algorithm is proposed to generate the desired predistorted signal to solve the nonlinear memory polynomial (NMP) linearizer's coefficients to realize the linearization. Moreover, in the processing of linearization, the NMP model only needs to be solved once, and hence, the linearizer could be implemented with low complexity. The experimental results verify that the normalized root mean square error (NMSE) performance of the NMP model can reach −39.3 dB and the NMP linearizer could improve the adjacent channel power leakage ratio (ACLR) performance by 16 dB when the PA is driven to the 31.3 dBm average output power with the strong nonlinear distortion.