Research And Hardware Implementation Of Wideband Digital Predistortion Based On Vector Quantization

In 4G and 5G mobile communications,high-order modulation methods are used in limited spectrum resources to achieve higher spectrum utilization.However,the increase of the Peak-to-Average Power Ratio(PAPR)of the high-order modulation signal will cause the non-linear distortion of the Power Amplifier(Power Amplifier,PA)to increase.Therefore,the linearization requirements of RF power amplifiers are also increasing,and Digital Predistortion(Digital Predistortion,DPD)is currently the most flexible and widely used linearization compensation technology.For broadband application scenarios,this paper applies Vector Quantization(VQ)algorithm to digital predistortion technology,and proposes Two-Dimensional Weighted K-means(Two Dimensional Weighted K-means,TDWK)algorithm and Constrained Two-Dimensional Weighted K-means(Constrained Two Dimensional Weighted K-means,CTDWK)algorithm,and simulated the modeling effect of the algorithm.In order to verify the actual compensation effect of the two algorithms proposed in this article on the power amplifier in a wideband scenario,this article designs a wideband digital predistortion hardware platform and verifies the feasibility of the algorithm through experiments.The main work content and innovations of this paper are as follows:This article first proposes an innovative switching behavior model,which uses a vector quantization algorithm to divide the input signal space into different regions,and uses a separate model corresponding to each region to calculate the output.Based on the K-means algorithm in the vector quantization algorithm,the TDWK algorithm is proposed.On the basis of TDWK,the prior knowledge of the size of the cluster is considered to improve the algorithm,and the CTDWK algorithm is proposed.The individual model of each region selects the General Memory Polynomial(GMP)model,combines these two algorithms with the GMP model,and proposes the TDWK-GMP model and the CTDWK-GMP model.The models involved in this article were used to model and simulate power amplifiers using signals with a bandwidth of 30 MHz,and the power amplifiers were selected as class F.The modeling simulation results show that compared with the GMP model,the modeling effect of the TDWK-GMP model is improved by 2d B,and the modeling effect of the CTDWK-GMP model is improved by6 d B.The sampling rate of the digital pre-distortion feedback channel needs to reach 5times the signal bandwidth,and the digital predistortion hardware experiment platform must meet this high requirement.Existing instrument platforms often cannot meet the requirements of too wide analysis bandwidth,which requires a digital predistortion platform suitable for broadband and even ultra-wideband.This article builds a digital predistortion hardware platform based on FPGA,AD9172 and ADC12DJ3200.The maximum analysis bandwidth supported by this platform is 1.5GHz.The experiment selects the 16 QAM signal of 135 MHz bandwidth as the input signal of the class F power amplifier,and uses the 4 kinds of models mentioned in this article to carry on the digital predistortion compensation comparative experiment to the power amplifier respectively.The ACPR compensation results show that,compared with the GMP model,the compensation effect of the TDWK-GMP model is improved by 3d B,and the compensation effect of the CTDWK-GMP model is improved by 5d B.Experiments have proved that the two algorithms mentioned in this article can be applied to the research of power amplifier digital predistortion compensation in broadband scenarios.