Key Technology Of Out-of-Band Digital Predistortion For Power Amplifier Near Saturation

Power amplifiers of modern wireless transmitters usually work near saturation to ensure high power and efficiency.However,the strong nonlinearity of the power amplifier near saturation will cause spectrum regrowth with reduced utilization of spectrum resources.Digital predistortion can handle the nonlinearity while maintaining the transmitting power.It enables the miniaturization of equipment and facilitates the reduction of weight.Conventional digital predistortion methods have a high degree of compensation for the power amplifier nonlinearity.For nonlinearity of near-saturation power amplifier,the predistorted signal will enter the saturation of the power amplifier and cannot effectively suppress the out-of-band distortion.Therefore,this dissertation explores key technologies of out-of-band digital predistortion for power amplifiers near saturation.The main achievements includes:First,the feasibility of out-of-band digital predistortion under the near-saturation nonlinearity is analyzed and verified.(1)A signal model for the out-of-band digital predistortion is developed,and the optimization objective of the out-of-band predistorted signal is specified.(2)The out-of-band predistorted signal is converted into the solution of a nonlinear equation.By using the Banach fixed-point theorem(also called the contraction mapping principle)to find the numerical solution of the equation,two offline iterative algorithms are proposed to generate the out-of-band predistorted signal,where the convergence condition of the iteration is demonstrated.(3)Experimental results and theoretical analysis based on the offline iterative algorithms show that,with the nearsaturation nonlinearity,conventional digital predistortion methods cannot effectively suppress the out-of-band distortion.In contrast,the out-of-band digital predistortion technique can improve the performance of the out-of-band distortion suppression by sacrificing the performance of the in-band distortion suppression.Therefore,it is feasible to use the out-of-band digital predistortion to effectively suppress the out-of-band distortion under the near-saturation nonlinearity.Second,cascade-type out-of-band digital predistortion structures are proposed.(1)By reconstructing the offline iterative algorithm,two cascade-type predistortion structures are proposed to generate out-of-band predistorted signals in an online process.The stability of the cascade-type predistortion is analyzed.(2)Cascade-type predistortion with improved model basis functions is investigated.A nonlinear model is proposed,whose basis functions can be flexibly constructed using different nonlinear functions.Therefore,by utilizing a nonlinear function with bounded output to construct the basis functions of the proposed model,the stability can be improved for the cascade-type predistortion structure.(3)Cascade-type predistortion with improved model identification is investigated.Three model identification methods are proposed according to the regularization of the least-squares estimation and the adjustment of training samples.They can further improve the stability of the cascade-type predistortion.(4)Experimental results demonstrate that the cascade-type predistortion structure can perform similar to the offline iterative algorithm.Third,an indirect-learning-type out-of-band digital predistortion method is investigated and derived.(1)According to the offline iterative algorithm,an out-of-band frequency selective nonlinear model is proposed to generate the out-of-band predistorted signal in an online process.This model has a non-cascade structure,and its complexity is lower than that of the cascade-type predistortion structure.(2)The conventional indirectlearning model identification is analyzed,and the reason why it cannot identify the outof-band frequency selective nonlinear model is given.By improving the generality of the conventional indirect-learning architecture,a generalized indirect-learning architecture is proposed for out-of-band frequency selective model identification.(3)By combining the out-of-band frequency selective nonlinear model with the generalized indirect-learning model identification,an indirect-learning-type out-of-band digital predistortion method is proposed.(4)Based on the complexity analysis and experimental verification,the engineering implementation is suggested for the out-of-band digital predistortion.If the focus of the engineering implementation is low complexity,the indirect-learning-type out-of-band digital predistortion should be considered.If the focus of the engineering implementation is high performance on the suppression of the out-of-band distortion,the cascade-type out-of-band digital predistortion should be considered.This dissertation studies the out-of-band digital predistortion for power amplifiers near saturation to suppress the out-of-band distortion.The achievements of this dissertation can can provide support for relaxing requirements on the analog filter of the radio frequency front-end.Therefore,the power loss of the radio frequency front-end can be decreased,and the miniaturization of equipment as well as the reduction of weight can be achieved.