Research On Digital Predistortion Technology For Hybrid Beamforming Architecture Transmitters

The Power Amplifier(PA)is an important active component in modern RF transmitters.The inherent non-linear distortion characteristics of PAs are the main source of RF impairing in wireless communication systems,which can increase signal error rate and cause interfere with the transmission of adjacent signals.Digital Predistortion(DPD)technology is currently the mainstream solution for linearizing PAs.With its excellent correction performance,high flexibility,easy implementation,and low cost,DPD has become an important part of modern wireless communication systems.As people’s demand for data communication continues to grow rapidly,communication systems are developing towards high carrier frequency,large signal bandwidth,low transmission delay,and so on.At the same time,in order to improve system capacity and transmission rate,Multiple-Input Multiple-Output(MIMO)technology and Beamforming technology are also being applied to communication systems.Therefore,this dissertation mainly focuses on the research of DPD technology for Hybrid Beamforming(HBF)architecture transmitters for future communication systems.HBF architecture transmitters are mainly used in massive MIMO systems that transmit high-frequency modulated signals.The spacing between the units in the array is extremely small.To reduce system complexity and hardware deployment costs,isolators are often no longer configured between the PA and the antenna.Due to the existence of crosstalk signals between the antennas,when the transmitter changes the beam direction to track the user’s position,the corresponding PA output impedance changes,which in turn changes the non-linear transmission characteristics of the transmitter,namely the load modulation effect.The predistorter needs to be continually activated and updated to ensure real-time linearization correction performance.This dissertation fully analyzes the impact mechanism of crosstalk signals on the main beam signal in different directions in a single-user scenario,and proposes a model that separates the crosstalk effect from the non-linear transmission characteristics of the PA based on Neural Network(NN).By analyzing the transmission characteristics of the transmitter in different directions,good performance can be ensured in all directions.The experimental results show that,compared with the model obtained from a fixed angle,the proposed model has better generalization performance in all directions,whether forward modeling or linearization correction.In the multi-user scenario,the Fully-Connected HBF(FC-HBF)architecture transmitter can transmit signals with higher gain,but the system also exhibits more complex transmission characteristics.Firstly,This dissertation analyzes its transmission characteristics.Unlike traditional transmitters that mainly transmit single-path signals,the PA in the FC-HBF system amplifies all user’s signals simultaneously,requiring simultaneous consideration of multiple inputs.Then,the dissertation introduces a polynomialbased model extension for multi-input models and points out its performance limitations.Finally,a model that combines the structural characteristics of the FC-HBF transmitter and Neural Network is proposed,which can simultaneously represent the transmission characteristics of the transmitter in multiple directions.Simulation results show that,compared with its good performance in weak non-linear distortion scenarios,the model based on polynomial extension has reached its performance limit in strong nonlinear distortion scenarios,and its linearization correction performance has decreased significantly.However,the proposed model has perfect correction performance in both scenarios.