Research On Digital Predistortion For Wideband Radio Frequency Power Amplifier

Accompanying with the applying of the 4^th generation mobile communication techniques,the bandwidth of the transmitted signal of wireless transmitter is wider and wider.This may cause that the output distortions of radio frequency?RF?power amplifier?PA?in the wireless transmitter are greatly increased.In order to reduce the influence of distortion on the quality of transmitted signals and interference to the adjacent channel cells,the wideband digital predistortion?DPD?techniques are applied to correct the output distortion of the radio frequency power amplifier.However,the wider bandwidth of transmitted signal,the higher peak to average power ratio?PAPR?and the bigger dynamic range of the transmitted signal make the cost of implementation,the resource of sampling and the practicability of system suffer more challenges.Consequently,we need focuses on analyzing the characteristics of the wideband RF PA,basic methods and learning structure of DPD technology.Further study DPD methods from the model structure implementation,the feedback sampling and the practicability of system,it can get more concise in model structure implementation,lower sampling rate in feedback and more stable in system application.The above issues are systematically and deeply investigated in this dissertation.The main research contents and contributions of this study are as follows:1.Aiming at the high complexity problem of model structure implementation,a neural network digital pre distortion model is structured,which use low order generalized memory polynomial basis to enhance the adaptation between the network model and the memory effect.Meanwhile,it adds the genetic algorithm to simplify the neural network model.The logic structure of look-up table method is realized based on linear interpolation to further reduce the logic resource.The experimental results show that the proposed method converges significantly faster,but also further enhance the linear performance,while the network scale has been controlled to be simplified,and logic resources is effective reduced.This method solves the high complexity problem of model structure implementation of the neural network model.2.The feedback under-sampling of DPD is intensively studied.How to reduce the feedback sampling rate is analyzed from four aspects.It includes:first,in order to overcome the difficulty of iteration when DPD uses feedback under-sampling,it establishes a digital pre distortion and power amplifier cascade model to complete the digital pre distortion under sampling iteration,while using non zero padding to recover the full rate of information,enhance the feedback under-sampling linearity correction performance.Second,aiming at the problem of under-sampled DPD linearity performance descended,it puts forward a model structure based on adaptive segmentation Lagrange basis,through direct learning structure and normalized least mean square error algorithm to train the parameters to improve the digital predistortion linearity correction performance,combined with adjustable fractional delay finite impulse response filter to recover the information of signal from full rate sampling.Third,in order to solve the image distortion problem of transmitter using under-sampling,an image distortion of transmitter correction method using feedback under-sampling is proposed.It can correct the distortion of the transmission channel and RF power amplifier distortion by the nonlinear model.The recursive least squares algorithm is used to speed up the convergence of the parameters.Fourth,for the high complexity problem of DPD parameter calculation,a frequency selective digital predistortion compression sampling optimization method is proposed,the frequency domain signal of predistortion is sparse sampled to reduce the computational complexity of the predistortion model.The experimental results show that the proposed scheme can solve the above-mentioned problems in the condition of reducing feedback the sampling rate of the digital predistortion.3.In order to solve the practical problems of the system under the condition of multi band and large dynamic signal,the following three aspects are analyzed from the aspects of system simplification,system robustness and system stability.First of all,for the problem that digital predistortion system simplifies in concurrent dual band transmitter,a hybrid designed structure in dual band digital predistortion system is proposed,it can reduce the model dimension to decrease the system resources.Then,aiming at the poor robustness problem in the vector switch model under the condition of some region sparse samples,an improved vector switch model is proposed,constrained least squares method is used to enhance the robustness of system application.Finally,for the problem that the large dynamic of the signal power caused poor practicability in digital pre-distortion system,a three-dimensional power segmented tracking adaptive digital pre distortion method is proposed,the long time average power,short time average power and instant power to build wideband RF PA thermal memory effect and short-term electrical memory effects tracking,which enhance the stability of the system application.The experimental results show that the proposed scheme can solve the existing practical problems of digital predistortion.