Research On Simplification Of Wideband Digital Predistortion Model Based On Compressed Sensing

Due to the inherent non-linear distortion characteristics of power amplifiers,the output will generate new frequency components.These new frequency components will cause distortion of the communication signal in the working channel.The distortion of the signal,which falls on adjacent channels,can interfere with the operation of other communication systems.Moreover,with the rapid development of the communications industry,high-efficiency modulation methods used to increase the communication rate make the signal have a high peak-to-average ratio,which further aggravates the nonlinear distortion of the power amplifier.In this context,the linearity of the power amplifier must be improved to ensure its efficiency.The thesis mainly studies the key technical problems of nonlinear correction of radio frequency power amplifiers in wideband wireless communication systems.Firstly,the thesis introduces the nonlinear distortion characteristics of radio frequency power amplifier and its related evaluation indexes in detail.The common nonlinear behavioral models of power amplifiers are studied.On this basis,the digital predistortion technique for power amplifier linearization correction is discussed.The simplification nonlinear behavior model of the power amplifier is emphatically studied.Combined with the greedy algorithm in the compressive sensing theory,a Sparity Adaptive Matching Pursuit(SAMP)clipping model based on Dice coefficient atom matching criterion is proposed,abbreviated as DSAMP sparse model,and compared with SAMP sparse model and Regularized Sparity Adaptive Matching Pursuit(RSAMP)sparse model.The simulations verify that the model has improved convergence speed and accuracy.Then,we focus on the parameter identification algorithm of power amplifier behavior model.Focused on the Newton algorithm and Quasi Newton(QN)algorithm based on the least mean square criterion,this thesis proposes a new improved Quasi-Newton.Improved Quasi Newton(IQN)algorithm,combined with Subspace Matching Pursuit algorithm based on Dice criterion in compressed sensing theory,studies a new power amplifier behavior model clipping algorithm,which is called DSMP-IQN algorithm.This adaptive sparse algorithm is applied to the clipping of Generalized Memory Polynomial(GMP)predistorter with third-order memory depth and tenth-order nonlinearity,and the simulation results of the different parameters and the sparsity are tested.It shows that the studied algorithm can effectively simplify the power amplifier predistorter model.The 148 coefficients of the Generalized Memory Polynomial predistorter model can be trimmed to less than 50%,and the comparable predistortion performance can be maintained.Finally,in order to further verify the validity of the proposed model,a digital pre-distortion verification system based on the instrument platform was built.The dual-band power amplifier based on harmonic control and Doherty power amplifier were selected as the test model.Then WCDMA and LTE signals with 10 M bandwidth are tested respectively as input stimulus.The experimental results show that the proposed new models have good performance in the power amplifier digital predistortion linearization system,which verifies the feasibility of the proposed models.