Research On Receiver-side Treatments For Nonlinear Distortion Of Power Amplifier

A crucial component at the transmitter of communication systems is the power amplifier(PA).Due to the inherent nonlinear property of PAs,nonlinearity will occur after the tansmitted signals pass through PA,which can degarde the communication quality of the linear system.With the development of mobile communication technologies,the base stations gradually present the trend of low cost and miniaturization,which also limits the transmission power of the signals.With the advantages of the high work efficiency and the significant ability to tackle the nonlinearity,the receiver-side technologies for combating the nonlinearity have attracted the attention of researchers.Firstly,this thesis introduces the nonlinear characteristics and the behavioral models of PAs.Then,on the basis of the memoryless polynomial model,two existing receiver-side techniques for tackling nonlinearity are studied;they are named as reconstruction of distorted signals(RODS)and power amplifier nonlinearity cancellation(PANC),respectively.Meanwhile,this thesis proposes two improved receiver-side methods,termed sum-division(SD)and sum-division with decision(SDD),respectively.Furthermore,this thesis also studies the schemes based on the neural network,and some improvements are given in the complex-valued neural network(CVNN)scheme.Finally,a new method is delivered,which combines the deep neural network with the SDD to combat the nonlinearity.The innovations and main works of this paper include the following aspects:1.The RODS and the PANC techniques are ananlyzed.Since they need to iteratively operate the received signals in the frequency domain,a lot of DFT and IDFT calculations are required,which leads to the increase of the computational complexity.In order to decrease the computational complexity,this thesis proposes the SD technique according to the characteristics of memoryless polynomial model,which attempts to reconstruct the transmitted signals in the time domain by calculating the nonlinear parameters.In comparison to RODS scheme,SD can achieve the same BER performance via a lower computational complexity.2.On the basis of the SD method,the SDD scheme is developed by adding the decision into the whole iterative calculation.In order to compare the BER performance of above methods,simulations are conducted via the MATLAB.The experimental results show that the SDD technique can outperform other methods under the given simulation conditions.3.Based on the idea of piecewise fitting,piecewise complex-valued neural networks(PCVNNs)is proposed to tackle the nonlinearity at the receiver.This method first divides the received signals into several intervals by their amplitudes,and each interval has a complex-valued neural network(CVNN)to process the signals.The received signals need to be mapped into the high-dimensional and high-order input vectors,and then it will be sent into the corresponding CVNN for the estimation of the transmitted signals.The simulation results verify the effectiveness of PSCVNNs scheme.4.This thesis presents a new approach to deal with the nonlinearity,which combines the deep neural network(DNN)with the SDD method.At first,the DNN is employed to reduce the amplitude distortion of the received signals,and then the output of the DNN is sent into the SDD module as the initial input.Finally,the estimation of the transmitted signals can be completed by the SDD.By optimizing the initial estimation,the iterations of the SDD module can be reduced,which can lead to a fast estimation of the transmitted signals.Finally,the experimental results are given by the MATLAB.