Research On Undersampling Digital Predistortion Technology

In the wireless communication system,the RF power amplifier is an indispensable key device of the transmitter.Its working performance has an important impact on the wireless communication system.This effect is mainly reflected in the energy loss and spectrum interference.The main reason is the power amplifier itself.Non-linear characteristics and memory effects.In modern wireless communication technologies(4G,5G),high-order modulation methods and larger signal bandwidths make power amplifiers exhibit stronger nonlinearity,which brings adverse effects to the communication system.Digital predistortion technology is one of the methods to effectively improve this problem.It is worth noting that in traditional predistortion systems,AD sampling rate of 5 times the signal bandwidth is required.As the signal bandwidth increases,digital predistortion systems require higher sampling.ADC(Analog-to-Digital Converter),which will cause the digital predistortion system to face hardware cost issues and energy consumption issues.In response to this problem,this paper focuses on the digital predistortion technology under feedback channel undersampling conditions(less than 5 times the signal bandwidth);because the behavior model of broadband power amplifiers has always been one of the research priorities in digital prediction technology,this paper also The weighted memory polynomial model is studied.Based on the memory polynomial model,the weighted memory polynomial introduces the input signal amplitude as a parameter of the model.According to the static nonlinearity and dynamic nonlinearity of the power amplifier,it shows two different characteristics with the change of the input signal amplitude.The model is divided into The static part and the dynamic part respectively use the static and dynamic weight functions to realize the different weights of the static and dynamic parts of the model.With the same number of model coefficients,weighted memory polynomials and memory polynomials have similar computational complexity.Using weighted memory polynomials can achieve better modeling results;when achieving the same modeling effect,using weighted memory polynomial models requires less Model coefficients.When the modeling error is-36 d B,the number of weighted memory polynomial model coefficients is 40% less than the memory polynomial model;when the same number of model coefficients,the weighted memory polynomial model can achieve a modeling effect of about-40 d B,while the memory polynomial model is used It can only reach about-36 d B.In addition,this paper studies the impact of the threshold value of the weight function on the model performance,and for this problem,the weighted memory polynomial model uses a weight function based on the cosine function,while maintaining the performance of the model while avoiding the use of the model due to the threshold band Instability.Digital prediction technology based on under-sampling restoration(USR,undersampling restoration)can realize digital prediction in the condition that the ADC sampling rate is less than 5 times the signal bandwidth.USR refers to a method of collecting the input and output signals of the power amplifier under under-sampling conditions and using the real power amplifier input signal to restore the real power amplifier output signal.In this method,the forward model of the power amplifier is established based on the restored power amplifier signal.By iteratively modifying the power amplifier model,a sufficiently accurate output signal of the power amplifier is finally restored.This paper introduces Landweber iterative algorithm on the basis of USR,and proposes digital predistortion based on USR-Landweber.USR-Landweber uses the USR-based digital predictive true structure and the iterative process is the same.The power amplifier data restoration uses the Landweber iterative algorithm.The data restored at the same sampling rate is closer to the real power amplifier output signal,which can be achieved.Better pre-distortion effect.Compared with the USR-based digital prediction,there is a certain improvement in performance.Finally,a digital pre-distortion experiment platform was set up with related instruments,using a LTE(Long Term Evolution)signal with a bandwidth of 5M as an input signal,and a Class F RF power amplifier was tested at a sampling rate of 10 Msps.The output signal ACPR is reduced from-30 d B to-47.04 d B before predistortion,and ACPR is further reduced to-48.84 d B after Landweber-USR algorithm predistortion.