Research Of Low Complexity Wideband Digital Predistortion Technology

Power Amplifier(PA)is an indispensable device in wireless communication system.Its performance directly affects the quality of the whole wireless communication system.But the core device of power amplifier is semiconductor amplifier,which has inherent non-linearity,so it is very important for the research of power amplifier linearization.Digital PreDistortion(DPD)has become a hotspot in the research of power amplifier linearization technology because of its simple implementation and easy debugging.Currently,it is in the transition period from 4G to 5G,and the fifth generation mobile communication is about to be commercially available.The requirements of "large bandwidth" and "small base station" in the fifth generation mobile communication bring new challenges to digital predistortion.The sampling rate of Analog-to-Digital Converter(ADC)in the feedback loop needs to be five times as wide as the original signal bandwidth in traditional digital predistortion.ADC is a key device with high cost and high power consumption in DPD system.Therefore,under the background of 5G wide bandwidth,the cost and power consumption of DPD module will be greatly increased if we continue to use traditional DPD method.To solve the problem of traditional digital predistortion in 5G,this thesis introduces two algorithms to reduce the ADC sampling rate and DPD system complexity to meet the needs of large bandwidth and small base stations in 5G.The spectral extrapolation method is introduced in the thesis to reduce the sampling rate of ADC,which tends to recover the full band PA output signal iteratively using the band limited signal.In this thesis,the Simultaneous Perturbation Stochastic Approximation(SPSA)algorithm is used in spectrum extrapolation algorithm to replace the least square(LS)method in the process of solving the spectrum extrapolation operator,which greatly reduce the complexity of the DPD system.In this way,the requirements for both large bandwidth and small base stations in 5G can be met.To further reduce the complexity of the DPD system,the Stagewise Weak Orthogonal Matching Pursuit(SWOMP)algorithm was used for power amplifier model clipping.Based on the SAMP and OMP algorithms,an improved SWOMP algorithm is proposed,which can crop the power amplifier model without knowing thesparsity.Finally,the validity of SWOMP algorithm is verified by comparing it with OMP algorithm and SAMP algorithm.In order to verify the corrective effect of the proposed algorithm,test and simulation are carried out based on the instrument platform.Class F power amplifier and Doherty power amplifier are used as the power amplifiers to be measured,and10 M LTE signal and 10 M WCDMA four carrier signal are used as the signal to be measured.The experimental results show that the proposed two algorithms have good digital predistortion effect when they meet the requirements of "large bandwidth" and "small base station".