Research On Broadband Digital Predistortion Technology And System With Optimized Basis Function And Quantization Accuracy

The development trend of broadband,miniaturization and high efficiency of mobile communication system puts forward higher requirements for the high efficiency and broad bandwidth output of radio frequency(RF)power amplifiers(PAs).The inherent non-linearity of the RF PAs will seriously deteriorate the quality of the transmitted signal and interfere with signals in the adjacent channels.As the dominant linearization methods for RF PAs,digital pre-distortion(DPD)evolves with the development of mobile communication system.Firstly,the application of 5G technology raises the requirements for the broadband DPD adapting to the massive MIMO architecture.The broad bandwidth has increased the complexity of digital predistortion model,and the development of intellectualization also requires selfadaptive DPD.In this case,DPD with adaptive optimization of high-performance and low complexity models is on urgent demand.In addition,the development of large bandwidth and miniaturization of communication system requires digital predistortion to ensure the linearization performance and reduce the hardware complexity.Therefore,this dissertation will focus on the above problems and study these three aspects of undersampled DPD under MIMO hybrid beamforming architecture,model adaptive optimization of wideband PA,and DPD system under low-complexity feedback link.Firstly,the non-linear behaviour of the array amplifiers’ non-linearity combined with the wireless channel is analysed on a massive MIMO system with hybrid beamforming.And a single user undersampling DPD architecture is proposed.The undersampling is achieved by introducing a band-limiting filter in the feedback loop to limit the sampling bandwidth,while an inverse beamforming module is set up to generate the equivalent signal of the main beam in the far field to participate in the extraction of DPD parameters.This is then further extended to a fully connected hybrid beamforming multi-user scenario,where a pre-distortion model for the multi-user scenario is deduced,and an undersampling DPD method for the multi-user scenario is proposed.In addition,the consistency assumption on array PAs in the above study is relaxed,and a low complexity array PAs inconsistency pre-compensation method is put forward.Simulation tests verify the capability of the proposed method to effectively reduce the sampling bandwidth and the significant compensation effect for array PAs inconsistency.Secondly,a DPD method with high efficiency and adaptive optimization of basis function is presented.A DPD model optimization mechanism is established to adaptively search for the most suitable DPD based on the current input signal and the operating state of the PA.Taking reducing the complexity of the model and improving the performance of forward modeling as the two main objectives,this dissertation constructs a suitable cost function.Then,two heuristic search algorithms with high robustness and strong search capability are analysed and proposed.The obtained algorithms both have strong global search capability and can quickly converge to the optimal solution set;at the same time,they exhibit strong local search capability in the optimal solution set,refining the search to select the optimal model structure with both high modeling accuracy and low complexity.The algorithm also considers the correlation between the basis functions of Volterra-based model,and uses the orthogonal matching pursuit algorithm to re-optimize the searched optimal model structure,so as to further reduce the adaptive complexity of the model.Experimental tests show that the algorithm has the ability to converge quickly to the optimal solution set,and that the optimised model has comparable performance to the conventional search model in both 60MHz and 100MHz scenarios,while reducing the computational complexity by approximately 75%and 80%,respectively.Finally,two DPD methods are proposed for reducing the dynamic range of the feedback link and for undersampling.In terms of reducing the dynamic range of the feedback link,a closed-loop digital pre-distortion architecture based on a feedback link with low-bit ADC is constructed and an update formula for the pre-distortion parameters is derived.Then,a realistic digital pre-distortion architecture adopting only one feedback loop is proposed via analysing the statistical characteristics of the signal.In terms of reducing the sampling bandwidth,a feedback link sampling bandwidth much lower than the amplifier output bandwidth is achieved by multi-narrowband spectral devision undersampling,and the complete PA output signal is then estimated by forward modeling of the subband signal after upsampling and digital frequency conversion for parameter extraction of the DPD.Experimental tests verify the effectiveness of the proposed method:compared with the ideal digital predistortion algorithm with high precision and full sampling,the above two methods effectively reduce the hardware cost of the feedback link on the premise of maintaining the linearization performance.In the closed-loop single channel low bit digital predistortion,only one ADC sampling feedback link with 8-bit quantization accuracy is used;In the spectral division sampling based on forward modeling,the digital predistortion method reduces the sampling bandwidth to about 1/6(and below)of the full sampling of the output signal.In conclusion,the research on the above three kinds of engineering problems effectively solves the problems of linearization of array PA,adaptive optimization of DPD model and high complexity of feedback link in ultra wideband scenario.Therefore,the undersampling DPD of MIMO hybrid beamforming architecture,the adaptive optimization algorithm of broadband PA model and the implementation method of DPD system under low complex feedback link proposed in this dissertation provide an important reference for the broadband,high efficient and low-cost miniaturization design of 5G system PA.