Research On The Digital Predistortion Algorithm Of Wideband Power Amplifiers Under The Framework Of Compressed Sensing

Wideband power amplifiers are critical components in present wireless communication systems,and their enhanced linearity is critical for communication quality and transmission efficiency.Higher-order modulations are commonly employed to improve spectral efficiency,however,it results in higher peak-to-average power ratios,which produces severe nonlinear behavior.The digital predistortion is a widely used linearization technology,which can well solve the problem of signal distortion caused by the nonlinearity of power amplifiers.However,as the bandwidth expands,the sampling rate in the digital predistortion feedback loop increases,making hardware implementation more complex.While,sampling and reconstruction may be done at a lower rate than Nyquist utilizing the compressed sensing principle.As a result,it is critical to investigate the digital predistortion algorithm of broadband power amplifiers within the framework of compressed sensing.A power amplifier behavior model based on the dynamic weighting function which combines the dynamic deviation reduction idea with the dynamic weighting function on the basis of the traditional Volterra series model and the generalized memory polynomial model is developed in consideration of the current polynomial power amplifier behavior model’s excessive complexity and a large number of coefficients.Experimental simulation structure shows the normalized mean squared error value is increased by 2-3d B when compared to the generalized memory polynomial and dynamic deviation reduction models.Then the value of adjacent channel power ratio is improved by 15 d Bc,and the number of model coefficients is greatly decreased when compared to the circumstance without predistortion.And the model can better simulate the memory effect of the power amplifier and the dynamic distortion of different intensities,reducing the model coefficient and complexity.To address the problem of poor linearity and complex modeling of power amplifiers,the application of compressed sensing algorithms in the pre-distortion modeling method is studied,and the frequency-domain notch based on related support set selection of sparsity adaptive matching pursuit(RSS-FNSAMP)algorithm is proposed.The framework is focused on the reconstruction algorithm of compressed sensing,which is sparsity adaptive matching pursuit algorithm,and used a related support coefficient matrix to select the optimal support set by regulating out-of-band distortion through frequency-domain notch.Simulation analysis shows that compared with orthogonal matching pursuit,subspace pursuit,sparsity adaptive matching pursuit,doubly orthogonal matching pursuit algorithm,the RSS-FNSAMP algorithm offers more reconstruction capability and can perform more accurate reconstruction in cases of high sparsity,lowering model complexity.A segmented dual-feedback pre-distorter structure based on the digital predistortion indirect learning structure is proposed to address the problem of too many system parameters being processed by the indirect learning structure.The structure obtains the optimal support set by constructing a negative feedback iterative loop,defines it to the pre-distorter in the forward path,then extracts the model coefficients by behavioral model identification algorithms,and finally copies the values to the pre-distorter to identify the optimal digital predistortion signal,eliminate modelling errors and perform parameter identification.The degree of improvement of the amplifier non-linearity by the proposed structure and the RSS-FNSAMP algorithm is verified by simulation experiments.Compared with the case without pre-distortion,the adjacent channel power ratio of the proposed algorithm in this paper is reduced by about 21 d Bc at the low and high frequency bands.Then compared with ILA-SAMP,ILA-DOMP and segmented double feedbackDOMP,the value of adjacent channel power ratio is improved by about,10 d Bc,7d Bc,5d Bc at the low frequency band and the high frequency band respectively,and the value of normalized mean squared error is improved by about 3-5d Bc.The paper has 47 figures,12 tables,and 63 references.