Study And Performance Analysis Of Hybrid Beamforming In Massive MIMO Systems

To cope with the conflicts between the rapid growth of future data transmission needs and limited radio resources,mobile communication systems have to further improve the spectral efficiency.Massive multiple-input multiple-output(MIMO),equipped with large-scale antenna arrays,has been one of the key technologies which is able to fully exploit spatial resources and remarkably improve spectral efficiency.In order to decrease the system power,hardware costs and complexity,the hybrid beamforming,where multiple antennas share fewer radio frequency chains,becomes one of the research hotspots in massive MIMO.The hybrid precoding consists of analog precoding and digital precoding.The analog precoding becomes the main challenge in studies due to the constant amplitude constraints,complicated circuit configurations and “frequency flat” characteristic.This dissertation investigates the adpative hybrid precoding strategies,dissipation and quantied precoding and hybrid precoding designs in orthogonal frequency division multiplexing(OFDM)systems.First,the hybrid precoding procedure is investigated and optimized which further indicates adaptive hybrid precoding method based on system factors.Under Rayleigh fading channels,the closed-form spectrel efficiency expressions for MRT/ZF-based hybrid precoding and pure analog precoding are derived.Comparing the spectral efficiency performance of hybrid precoding and pure analog precoding,it gives the system factor conditions on the performance tradeoff between two precoding methods.According to the analyzed conditions,the adaptive hybrid precoding procedure is given to optimize the system performance.Moreover,the conclusions are also extened to uplink hybrid detection which are the same with those in the downlink systems.The numerical results show that the conclusions derived in Rayleigh fading channels also suit well in spatially correlated channels and 5G simulation channels.Then,considering dissipation in analog component network and effect of quantized precoding,the spectral efficiency performance of sub-connected and fully-connected structures are analyzed.Moreover,the performance tradeoff between the two structures are investigated.The analog precoding part in hybrid precoding is implemented by analog compenent network which leads to dissipation caused by a large amount of power dividers and combiners.We introduce the two typical circuit configurations of hybrid precoding systems,namely the sub-connected and fully-connected structures,by considering the dissipation and analyze the impact of the analog component network on signal power.Different from other existing researches,it is discovered that the sub-connected structure could obtain better spectral efficiency performance and lower complexity in comparison with the fully-connected structure on some system assumptions.Moreover,even if the physical channels are independent,the effective channels through analog precoding become correlated.Therefore,the hybrid system should utilize the channel correlation-based codebook which is proved to perform better than the conventional random vector quantization(RVQ)codebook.Finally,one alternating maximization framework for hybrid precoding with two iterative algorithms and one closed-form design are proposed for OFDM-based systems.In detail,the analog precoding is optimized using the conjugate gradient algorithm on the Riemannian manifold and the digital precoding is designed by a second-order cone program(SOCP)-based method which implies a locally optimal alternating maximization hybrid precoding design.Then,replacing the SOCP-based method by a weighted minimum mean square error(MMSE)-based strategy yields to another alternating maximization hybrid precoding design.In particular,it is proved that two alternating maximization algorithms have similar performance when the weights among all users are similar.Moreover,a closed-form hybrid precoding is also proposed which calculates the analog precoding utlizing channel matrix decompositon design(CMDD)and designs the digital precoding as a weighted MMSE precoding.Although the closed-form algorithm has some performance loss comapred to iterative algorithms,it requires lower system computational complexity.The closed-form strategy could be a good inilization for iterative methods,saving thereby itrations.Simulation results show that all the three proposed hybrid precoding methods for frequency selective have great performance.