Research On Efficient Channel Reconstruction Algorithms Based On Prefaded Signal Synthesis Method

In recent years,with the vigorous promotion of 5G commercialization,the market share of 5G massive multiple input multiple output(MIMO)device has gradually increased.Meanwhile,massive MIMO technology poses new challenges to the testing of 5G equipment,which further restricts the promotion and mass production of 5G equipment.Thus,multi-probe anechoic chamber(MPAC),as the MIMO device testing method which can theoretically reproduce any channel models,has become the focus of research in the industry.The main idea of prefaded signal synthesis(PFS),which is known as the significiant channel reconstruction method in MPAC,is to use limited number of probes to reproduce the disired spatial profiles within test area.However,due to the new characteristics of 5G MIMO devices,there are serveral challenges in the channel reconstruction of PFS method.Thus,in this paper,the PFS method is studied from the perspective of efficiency and accuracy.Firstly,one of the significant challenges of MPAC is the system cost reduction.Each probe in the testing system needs to connect with the expensive channel emulator(CE)port,whereas a large number of probes will increase the system cost greatly.Thus,the probe selection process is needed to reduce the number of probes in the testing system.However,in 5G MIMO OTA testing,due to the increase of candicate probes,the computation complexity of probe selection algorithm increases greatly,which seriously affects the efficiency of channel reconstruction of PFS method.This paper proposes an effective probe selection algorithm,which can effectively decrease the computation complexity compared with the mainstream convex algorithms.The main idea of proposed algorithm is to compute the gradient of objective function,and then adopt the momentum gradient descent method to optimize the probe weights.In order to select desired probes during iteration,the regularization function is added into the objective function.By dynamically adjusting the parameters in the iteration process,the accuracy and convergence speed of the algorithm are guaranteed.Through the theory analysis and simulation comparsion,it is found that the proposed algorithm has less complexity and higher accuracy,and can greatly improve the efficiency of channel reconstruction for PFS method.Secondly,the purpose of the PFSmethod is to accurately reproduce the spatial correlation of the target channel within the test area.In the previous literature,the deviation of spatial correlation within the test area is based on ensemble-averaging.However,in practical OTA testing,since every random parameter in the channel emulator is preconfigured,the every random parametor is preconfigured,the channel time response generated by the channel emulator is deterministic.Therefore,the hypothesis based on ensemble-averaging in the prevoius literature is not valid,resulting in errors in the simulated spatial correlation within the test area.In this paper,the space-time joint correlation function(STCF)of the target channel and OTA channel is derived and analyzed theoretically from the new perspective of time-averaging.The derivation results show that the STCF is determined by the initial phase of subpath in the channel model.According to the theoretical derivation results,an improved channel moodelling method based on discrete Fourier transform(DFT)matrix is proposed in this paper,which can eliminate the error of spatial correlation synthesis in the traditional PFS method.The main idea of proposed method is to use the orthogonality of DFT matrix and take each column vector in the matrix as the initial phase of subpath in the channel model.Simulation results show that the proposed method can greatly improve the accuracy of channel reconstruction.