Research On Low-complexity Digital Pre-distortion Model Based On RF Power Amplifie

In early wireless communication,constant envelope modulation is often used in the communication process.The advantage of this mode is that the power amplifier can work in the linear amplification area and improve transmission efficiency.However,the current spectrum resources are very scarce.In order to improve the utilization of the spectrum,modern communication has adopted digital modulation methods.This modulation method has the characteristics of a wide frequency spectrum,non-constant envelope,and peak-to-average ratio,resulting in power amplifiers working in areas near saturation points with very poor linearity.After passing through the power amplifier,the modulated signal will also be highly distorted.At the same time,because the input signals have different envelopes,the memory effect will also cause distortion,so it is very important to improve the linearity of the power amplifier.At present,the most widely used linearization technology is digital pre-distortion technology.Compared with other linearization technologies,digital pre-distortion technology has advantages such as broadband applicability,strong linearization ability,and low cost.The core of digital pre-distortion technology is to construct pre-distortion models with good linearization performance and low complexity and to study more effective parameter extraction algorithms.The main work content of this paper is as follows.Firstly,when low parameter configurations are used for pre-distortion models,the Hessian matrix in the Gaussian Newton method used for model parameter extraction may exhibit pathological issues,which affect the convergence speed and accuracy of the model.This paper proposes a parameter extraction algorithm based on the improved Gaussian Newton method.This method refers to ridge regression to alleviate the pathological problems in the least squares method,inserts a diagonal term matrix into the Hessian matrix,and adaptively adjusts the matrix coefficients according to the system error,thereby alleviating the pathological problems in the Gaussian Newton method.The simulation experiment results show that the convergence speed is significantly improved compared to the Gaussian Newton method.Then,in response to the problem of high coefficients and complexity in improving the nonlinearity of power amplifiers in current pre-distortion models,this paper proposes a predistortion model based on a dynamic generalized memory polynomial.This model introduces the cross term of the generalized memory polynomial model based on the dynamic memory polynomial model and combines dynamic thinking with the interaction between different memory moments.The experimental simulation results show that the proposed model can reduce most coefficients while maintaining high accuracy.Finally,based on the RF Web Lab testing platform,a digital pre-distortion technology testing process was developed for the linearization ability of the model and parameter extraction algorithm proposed in the paper in real testing environments.Firstly,experimental analysis was conducted on the improved Gaussian Newton parameter extraction method in a real testing environment.Then it was combined with a dynamic generalized memory polynomial model for comprehensive experiments.The experimental results show that the proposed method reduces the power ratio in adjacent channels by 3.9 d Bc and 6 d Bc,compared to the power amplifier without pre-distortion.Compared with the generalized memory polynomial model,the accuracy is the same,but the number of model families is reduced by 27.4%.