Filter-free Band-limited Digital Predistortion of Power Amplifiers for 5G Wireless Transmitters

Abstract

The increases of signal bandwidth in wireless communication systems pose new challenges to the digital predistortion (DPD) for the linearization of power amplifiers (PAs). To reduce the baseband bandwidth in the transmitter chain and the feedback sampling rate in the feedback loop, a band-limited DPD (BL-DPD) was proposed to linear the broadband PA output when sampling rates are insufficient. Most of the methods are based on the BL-DPD that adopts an analog filter in the feedback loop to limit the baseband bandwidth and feedback sampling rate of the DPD. However, the analog filter leads to the PA’s strong memory effect to increase the number of model coefficients of the DPD. In this paper, a novel filter-free band-limited DPD (FFBL-DPD) method, which combines oversampling method with an undersampling method to alleviate the PA’s nonlinear behavior with weak memory effect model, is proposed for the linearization of the broadband PA. First, a low speed analog-to-digital converter (ADC) will capture the PA output to obtain the low sampling rate PA output signal. Then, the DPD model with low sampling rate will be identified and constructed with the captured low sampling rate PA output signal and the original signal. It will be used to generate a low sampling rate predistorted signal. Theoretical analysis shows that most of the nonlinear compensation terms, which the conventional full sampling rate DPD model can obtain, can be generated by employing the DPD model with low sampling rate. The FFBL-DPD method can compromise a small part of the linearization performance. So it can significantly reduce the bandwidth requirements of the baseband, transmitter chain and feedback loop and the ADC/digital-to-analog converter (DAC) sampling rate. Finally, the low sampling rate predistorted signal will be oversampled to feed into the PA. To validate the proposed method, the simulation experiments are carried out with a 100-MHz OF DM signal. Experimental results show that excellent linearization performance can be obtained with fewer parameters at the limited available sampling rate than the existing band-limited DPD. Besides, it presents much lower complexity that will reduce expenses on hardware, which is very suitable for 5G broadband applications.