Low Computational Complexity Digital Predistortion Based On Independent Parameters Estimation

A power amplifier(PA)is an important active device in wireless transmitters and it is also the major source of nonlinear distortion in transmitters.In order to compensate the nonlinearity of radio frequency PA,a variety of linearization techniques have been proposed.Especially,Digital Predistortion(DPD)has become the mainstream linearization technique owing to its advantages such as the flexibility in programming,the relatively low cost,and the high accuracy.The current development trend of DPD focuses on low complexity DPD technology in wideband/multi-band applications.This paper focuses on the parameter identification of complex DPD models and the numerical instability problem caused by the high complexity has been sloved.This paper also proposes the implementation of low computational complexity DPD algorithms with lower power consumption and prices.In wideband applications,the number of coefficients in conventional models increases prominently to compensate the high nonlinearity and deep memory effect of the PA.And the increase of the coefficients leads to the negative impact on the process of extracting the DPD parameters and also leads to the numerical instability problem.Based on the principal component analysis(PCA)of the conventional DPD system,a low computational complexity DPD system has been proposed in this paper.Experimental results demonstrate that the proposed approach can reduce the number of coefficients of DPD models and prevent ill-conditioning with lower computational complexity.In concurrent dual-band applications,behavioral models take the form of a dual-input-single-output system with more coefficients,which need to be simplified.This paper extends the low computational complexity approach based on principal component analysis to the dual-band applications,and then derives the iterative formulas and condition of convergence.Furthermore,this paper analyzes the impact of the new basis functions on the performance of DPD models and proposes a new basis function selection strategy.This strategy considers the characteristics of PA in order to select basis function matrices with better performance.Different DPD model simplification methods directly affect the linearization performance.In order to further reduce the number of coefficients in DPD models with almost negligible performance loss,this paper proposes a novel DPD method based on partial least squares(PLS).This method obtains new basis function matrices based on the PA's inverse input and output characteristics to achieve a simplified model.Finally,experimental results demonstrate that the new method can greatly reduce the number of coefficients with almost no loss of performance.