Research On Adaptive Predistortion Technology For Broadband And High Capacity MIMO System

In recent years,with the rapid development of the mobile Internet and the internet of things,then promoted the growing development of wireless communication systems with high quality,low energy consumption,high speed,and large bandwidth.It brings a variety of new challenges to adaptive predistortion technology for power amplifer(PA): high-quality communication brings requirements for high performances,low energy consumption brings requirements for low complexity,and dynamic complex statistics characteristics of wideband signal need for stable identification capabilities.Aiming at above challenges,this thesis takes high-precision,strong-adaptability and stable multiple input multiple output(MIMO)digital predistortion technology as the research goal.The main work of this thesis is summarized as follows:First,this thesis introduces the basic theoretical knowledge of adaptive predistortion technology for wireless communication,then analyzes MIMO technology,which widely used in wireless communication systems.Then we conducts in-depth analysis and summary of the challenges brought by large-bandwidth and high-capacity wireless signals to classic digital predistortion technology.Based on the challenges,an adaptive predistortion scheme for broadband and high-capacity MIMO systems is proposed.The key technologies of the proposed scheme are developed around two issues: Firstly,in MIMO system full-loop interference is difficult to effectively suppress,which consists of crosstalks between multiple branches,quadrature modulator(QM)error,quadrature demodulator(QDM)error and power amplifie distortion.Secondly,PA predistortion is difficult to effectively done,which leaded by the dynamic complex statistical characteristics of broad-band high-capacity signals.Subsequently,aiming at the problem that the MIMO system's full loop interference is difficult to suppress,we study the high-performance,low-complexity MIMO adaptive predistortion architecture for joint full loop interference suppression.We analyze the reasons for the limited performance of the classical MIMO predistortion architecture and we propose a joint full-loop interference suppression MIMO adaptive predistortion architecture.In the first stage,jointly eliminate the QM error,nonlinear crosstalk,linear crosstalk,and QDM error.In the second stage,the predistortion processing is done separately for the PA distortion in each branch.The proposed MIMO predistortion architecture solves the problem that the full loop interference is too complicated to suppress effectively.Theoretical analysis and simulation results show that the proposed MIMO adaptive predistortion architecture has the advantages of robustness against suppress,high performance,and low complexity to classical MIMO predistortion architecture.Finally,aiming at the strong memory effect and nonlinear distortion of the power amplifier caused by the complex statistical characteristics of the broadband signal is too difficult to effectively compensate for,and we research the high-efficiency predistortion identification methods based on neural network.Through theoretical analysis,it is found that the classical predistortion identification methods and predistortion identification methods based on realvalue time delay neural network have limited compensation performances.Then this thesis proposes the predistortion identification methods based on improved real-value time delay neural network.The predistortion identification methods based on IRVTDNN optimizies input structure of the neural network,which starts from the match of the neural network and the power amplifier model.On the basis of the I component and the Q component,the amplitude and sine value of the input signal are added into neural network in the IRVTDNN,so the proposed predistortion identification methods has higher identification accuracy and stronger stability.Theoretical analysis and experimental simulation show that the predistortion identification methods based on IRVTDNN improves the system performance of linearization compensation at the expense of the less computational complexity.