Research On Linearization Technology Of S-band Harmonic Radar Power Amplifier Based On Digital Predistortion

Power amplifier is an important component in S-band harmonic radar transmitting system.Due to the nonlinear characteristics and memory effect of power amplifier itself,the signal generated by baseband will be distorted after passing through the power amplifier,which seriously affects the reception of the signal.Therefore,solving the problem of nonlinear distortion in power amplifier has become a new hotspot.Among them,the advantages of digital pre-distortion technology,such as strong adaptability and adjustable ability,have become the hot spot of power amplifier linearization technology.Based on the characteristics of S-band harmonic radar and the direct learning structure of predistortion,the traditional adaptive algorithm is aimed at the factors such as slow convergence speed,weak anti-noise ability,instability near the optimal value and easy to produce large deviation,which will lead to problems such as failure of convergence or large model error of the predistortion system.Therefore,based on the NLMS algorithm,this paper proposes an optimized NVNLMS algorithm.By changing the iteration function,increasing the autocorrelation matrix of adjacent errors,and adding the adjustment term,the convergence degree,anti-noise performance and steady-state error of the predistortion system are optimized.The experimental simulation results show that NVNLMS algorithm based predistortion system Normalized Mean Square Error than other two algorithms based on the system of 9.7 d B and 5.8 d B was optimized,respectively,under different environmental noise NVNLMS based digital predistortion system can fast convergence and compared with other two algorithm has good anti-interference and antinoise performance.The memory polynomial model based on Volterra series simplification is widely used in the establishment of predistortion model,but for the power amplifier with strong nonlinear degree and memory effect,it is difficult to identify accurately,or it is difficult to converge in the process of recognition with high computational complexity.Therefore,a Hammerstein model based on orthogonal memory polynomials is proposed.This model cascades memoryless nonlinear function modules with orthogonal memory polynomials modules,which can reduce the complexity of operation and accurately describe the nonlinear behavior and memory effect.The simulation results show that when the nonlinear order is 6,the normalized mean square error of Hammerstein model based on orthogonal memory polynomial is reduced by 4.7 d B and the number of model coefficients is reduced by 59.3%.Finally,NVNLMS algorithm and Hammerstein model of orthogonal memory polynomials are placed in the predistortion direct learning structure respectively,and compared with the other two predistortion system schemes.The simulation results show that the ACPR values of the predistortion system based on the proposed scheme are optimized by 9.1 d B and 6.9 d B compared with the other two predistortion schemes.Compared with the other two schemes,the EVM values based on digital predistortion of the proposed scheme are optimized by 1.89% and 0.61%.It is proved that this scheme can not only improve the pre-distortion effect,but also improve the anti-noise performance of the pre-distortion system.