Research On Linearization Technology Of Power Amplifier Based On Volterra Series

With the rapid development of modern mobile communications,spectrum resources and transmission channels in wireless communications are becoming increasingly tense,and green communication concepts that promote high-efficiency,low-energy-consumption and low-cost communication methods will be widely valued and vigorously promoted.Radio frequency power amplifier(Power Amplifier,PA)as the core component in modern mobile communication system,the improvement of its linearity has a significant impact on improving the communication quality,transmission efficiency and reducing energy consumption and cost of the entire communication system.Generally,feedforward,negative feedback,envelope elimination and recovery,and predistortion were used in communication systems to improve the linearity of power amplifiers.Among them,digital predistortion(DPD)technology has the most development prospect and engineering application value.As an important branch of digital predistortion technology,power amplifier modeling and predistortion methods based on Volterra series are due to their excellent linearization performance,supporting for broadband signal,simple structure and high stability,have been widely used in the linearization of RF power amplifiers.Combined with the current research status at present,this thesis studied the linearization performance and characteristics of the main models of RF power amplifiers based on Volterra series,and introduced a new output compensation memory polynomial(OCMP)model from the classic Volterra series.In addition,the digital predistortion learning structure was studied in depth,and the iterative learning control structure was introduced to further simplify and improve the linearization performance of the predistortion system based on the Volterra series.1.In view of the high complexity of the existing Volterra series model for power amplifiers and the difficulties in model coefficients extracting,this thesis focuses on the memory effect of the power amplifier from the traditional Volterra series,on this basis,the output cross term was introduced and a new output compensation memory polynomial(OCMP)model was proposed.In this thesis,the proposed model was verified and analyzed by simulation and experiment.In the experiment,a class F gallium nitride power amplifier and a Doherty power amplifier were linearized,and the output signal ACPR was reduced to below-52 dBc.Compared with MP model,DDR model and GMP model,OCMP model has obvious advantages in model complexity.2.By analyzeing the characteristics of the two main implementation architectures of the digital predistortion system: direct learning architecture(DLA)and indirect learning architecture(ILA),this thesis introduced an iterative learning control(ILC)architecture,and effectively combined the ILC structure with the Volterra series model of the power amplifier.This method simplifies the complexity of model parameter identification,and the ACPR value of the ILC learning architecture was improved by nearly 1 dBc compared with the ILA learning architecture.3.This thesis compared and analyzed the learning algorithm of the iterative learning control(ILC)architecture.Because the instantaneous gain ILC algorithm was greatly affected by system noise,the linear ILC algorithm used constant gain,the ability to model nonlinear amplifiers with strong nonlinear characteristics is limited,a piecewise linear ILC algorithm was proposed,which can accurately model power amplifiers with strong nonlinear characteristics,and at the same time has strong robustness to system noise.