Research On The Performance Limit Of Full-duplex Communication

In-Band Full-Duplex(IBFD)technology has the potential to double the spectrum efficiency without adding additional spectrum.This is especially attractive today when spectrum resources are increasingly scarce and the base of spectrum efficiency is already high.The biggest challenge faced by full-duplex technology is to overcome very strong selfinterference: Taking a typical Wi-Fi system as an example,the power difference between self-interference and background noise is as high as 110 d B.In order not to affect the performance of the system receiver,this requires the self-interference power to be reduced by at least 110 d B.Considering the various distortions(including linearity and non-linearity)and device noise that the self-interference signal suffers before reaching the receiving link,this is obviously a challenging task.Although existing research work has shown that 110 d B of self-interference cancellation can be achieved in a Wi-Fi system,however,whether full-duplex technology can further improve performance for use in wireless communication systems with a larger coverage radius(meaning higher requirements for the ability to eliminate self-interference)is currently unknown.Therefore,it is very important to probe and discourse the limit capability of self-interference cancellation of wireless full-duplex systems,which is the core goal of this thesis.Specifically,the main work of this thesis includes:(1)Probe and portrayed the upper limit of self-interference cancellation performance in the analog domain,and optimize the signal reconstruction scheme: We classify the analog domain self-interference cancellation scheme into two categories: signal reconstruction scheme and channel reconstruction scheme.In addition,the sources of reconstruction error are classified into three items: channel truncation error,coefficient estimation error,and fitting error.For the two types of reconstruction schemes,we analyzed the respective errors one by one,and described the upper limit of their performance.In addition,for the signal reconstruction scheme,this thesis proposes an interpolation method based on LS(least squares).Compared with the existing interpolation method(based on Sinc function interpolation),our scheme can obtain a very considerable performance gain;(2)Probe and portrayed the upper limit of self-interference cancellation performance in the digital domain,and optimize the design of channel training sequence: First,this thesis theoretically analyzes the simplified expression of residual self-interference power when using LS(least squares)method for channel reconstruction(including linear and non-linear channel components);On this basis,further theoretical analysis in this thesis shows that the residual self-interference power mainly depends on the peak-to-average ratio(PAPR)of the channel training sequence-the larger the peak-to-average ratio,the smaller the residual self-interference power.Therefore,this thesis proposes to use a random sequence based on modulation symbols as the training sequence.The simulation results show that this sequence has a very significant performance improvement over the existing training sequence based on the Gold sequence.In addition,the tests in this article show that the BPSK modulation sequence has a peak-to-average ratio comparable to that of high-order modulation sequences such as 256 QAM.Therefore,for the sake of simplicity,we propose to use BPSK random sequence as the training sequence for selfinterference channel reconstruction.The work of this thesis helps to deepen the understanding of the performance limits of full-duplex technology and guide the application of full-duplex technology in actual systems.