摘要

A novel direct-learning digital predistortion (DPD) of power amplifier (PA) is proposed to alleviate the requirement of an analog-to-digital converter (ADC) in the feedback path. Different from existing undersampling methods, the proposed approach requires only a single ADC running at low sampling rate, which saves one-half hardware cost and avoids the I/Q imbalance problem of the conventional zero-IF feedback structure. First, a modified Gauss-Newton iteration using a low-rate single-branch feedback signal is derived, which justifies the use of single undersampling ADC. Second, a low-complexity time and phase synchronization algorithm based on pattern cross correlation is presented, to align the input and output samples. Experimental results on a 10-W class-AB wideband PA and 100-MHz LTE-A test signal show that the proposed approach can effectively linearize the PA and outperforms other undersampling DPD approaches using the conventional zero-IF feedback structure, in terms of adjacent channel leakage ratio.