Research On Key Technology Of Hybrid Precoding Architecture Based Mmwave Wireless Transmission

A millimeter-wave(mmWave)communication system plays a promising role in future fifth generation cellular networks.The mmWave band can offer larger bandwidth communication channels than currently used bands in commercial wireless systems,however,the penetration losses are larger on the mmWave than on the lower-frequency wave.Therefore,a large antenna array with highly directional transmission/ reception should be made to compensate for the high penetration losses.However,hardware complexity and power consumption are large because of the use of large antenna arrays.Several architectures have been proposed to solve the hardware constraints in mmWave communication,including the hybrid analog/digital precoding combining architecture using phase array or lens and the low resolution analog and digital converter(ADC)architecture.Firstly,we study the hybrid precoding architecture based mmWave techniques.We analysis the transmit property of the mmWave,and the commonly used channel model for mmWave.Besides,we study the system architecture of mmWave multiple-input multiple-output(MIMO),such as the analog beamforming architecture,the hybrid precoding architecture,and the low-resolution ADC architecture.Then,we exploit the classical beamforming algorithms,including codebooks design,in-package beam training,and beam coding algorithms.Finally,we research the compressed sensing based channel estimation algorithms for mmWave.We compare the strengthes and weaknesses of the orthogonal matching pursuit(OMP)algorithm,the support detection(SD)algorithm,and the approximate message passing(AMP)algorithm.Subsequently,we design a low-resolution ADC module assisted hybrid beamforming architecture for mmWave communications.The proposed system architecture has a low-resolution ADC and a hybrid beamforming module.An electronic switch on the transceivers sweeps between the low-resolution ADC module and the hybrid beamforming module.During the training phase,the transmitter switches to the hybrid beamforming module and the receiver switches to the low-resolution ADC module.The electronic switch on the transceivers switches to the hybrid beamforming module during the data transmission phase.In addition,we design a fast beam training method which is suitable for the proposed system architecture.We divide the beam training process into two phases: in phase 1(All-Directions-Transmitting),the mobile station(MS)calculates all the received beam directions.In phase 2(Fine-Directions-Matching),the transmitted beams of the base station(BS)and the received beams of the MS are matched one-by-one.The simulation results show that the proposed hardware system architecture can successfully accelerate the millimeter-wave link establishment without degradation in the data transmission performance.The proposed beam training method requires only L + 1(where L is the number of paths)time slots which is smaller compared to the state-of-the-art.Finally,We formulate the channel estimation problem in the three dimension(3D)lens antenna array that can incorporate the sparse non-informative parameter estimator based cosparse analysis AMP for imaging(SCAMPI)algorithm to estimate the channel.We show that the SCAMPI algorithm can perform more effectively than other existing algorithms.To the best of our knowledge,this paper is the first study that demonstrates the channel estimation bridging to an image reconstruction technique.Besides,the value of each pixel in an image basically obeys uniform distribution.However,we observe that the channel responses follow nearly sparse Gaussian distribution in real life.Therefore,we replace the uniform prior distribution of channel responses with the sparse Gaussian distribution.The replacement introduces a practical challenge because several unknown parameters(such as sparsity rate,mean,and variance)appear in sparse Gaussian priori probability distribution function.Thus,we introduce the expectation maximization(EM)learning algorithm to find the(locally)maximum likelihood estimates of the parameters.The EM learning of the parameters is then deduced.The result reveals that,with the help of the EM learning algorithm,the sparse Gaussian is more suitable to be priori probability distribution than the uniform distribution used in images.Moreover,we introduce a new phase-shifter reduced measurement matrix structure in which a random part of the phase shifter is disconnected from the entire network.The new structure can reduce the power consumption of the system.We analyze the effect of the measurement matrix structure on the performance of the SCAMPI algorithm.The simulation results show that the SCAMPI algorithm is robust even if the number of phase shifters is reduced by 10%.