Research On Channel Sounding Methods And Spatial Correlation For Large-scale Antenna Systems

Wireless mobile communications have changed everybody's life and working styles profoundly. The mobile traffic will grow explosively with the rise of mobile internet and internet of things (IoT), which is a great challenge for the fifth generation mobile com-munication systems (5G). Massive multiple-input multiple-output (MIMO) is one of the key enabling technologies for 5G. It employs tens or even hundreds of antennas in the base station (BS), serving dozens of users with the same time and frequency resources.Massive MIMO has the advantages of extremely high spectrum efficiency, energy effi-ciency and reliability, which is capable of meeting the demand of huge mobile traffic in the future.Wireless channel characteristics are vital for system design and performance evalua-tion in massive MIMO systems. Channel estimation, antenna selection and channel state information feedback algorithms could be optimized based on antenna correlation in the array, and the complexity of pre-coding schemes depends on inter-user channel correla-tion. Therefore, it is important to investigate the antenna and user channel correlation in massive MIMO. Although lots of research on channel correlation has been performed worldwide, there are still some limitations as follows:1) For the channel sounding systems and scenarios, current channel measurements are mostly conducted with commercial channel sounders or related instruments, such as signal generators, signal analyzers and vector network analyzers. The high cost restricts extensive channel measurement campaigns. Besides, most channel measurements are conducted in outdoor scenarios, while indoor hot-spot is not fully investigated.2) For correlated fading models over large-scale antenna arrays, conventional an-tenna correlation models are still employed in optimization algorithm design for massive MIMO. These models are derived based on plane wave assumption, and their applicability should be revalidated. Further investigations and modelling on channel spatial continuity and randomness over the array in line-of-sight (LOS) scenarios are needed as well.3) For inter-user channel correlation characterization, eigenvalue spread and con-dition number of the multi-user channel matrix both have limitations, and the relation between user channel orthogonality and system setup has not been modeled mathemati-cally.This dissertation jointly employs channel sounding, simulation and mathematical modelling to investigate the channel spatial correlation characteristics in real massive MIMO environments. The main work and contributions of the thesis are listed as follows:1) A wideband channel sounding scheme with multiple frequency bands is proposed,and the design, implementation, calibration and verification procedures are detailed. In this scheme, multiple neighboring operation bands are sounded sequentially in one cycle to extend the overall bandwidth. The time consumption could be reduced significantly compared with the frequency-sweeping method, which is especially suitable for low mo-bility scenarios. As hardwares with very wide operation bandwidth are not necessary in this method, it reduces the cost and implementation complexity significantly. The sounder is implemented on a software-defined radio platform, and the impacts of non-ideal hard-ware could be removed with the proposed calibration method. The performance of the sounder has been validated through multi-path channel simulation as well as wired and wireless channel measurements.2) An auto-regressive moving average model (ARMA) is utilized to model the corre-lated fading process over linear large-scale antenna arrays. The ARMA model for Ricean K factors is further employed to improve the correlation-based stochastic channel mod-els (CSCM) in LOS scenarios. In the thesis, the variations of received signal power and Ricean K factors over the array are modeled as ARMA processes, respectively. Based on the ARMA model, an improved CSCM is proposed with varied K factors, simulating the relative power variation between the LOS signal and other multi-path components. This model is able to reflect the spatial continuity and randomness of the wireless channel over the array.3) Antenna correlation coefficients fitting and modelling over large-scale antenna arrays. The applicability of different existing antenna correlation models is verified for different spatial dimensions and room sizes. In vertical plane, correlation coefficients are affected by the antenna's radiation pattern and the environment,and a significant oscillatory decay pattern is observed. A damped cosine wave model is proposed in this condition, which achieves good performance.4) Asymptotic user channel orthogonality is analyzed and an empirical formula is built based on measurement results. The upper bound of the sum rate capacity in multi-user massive MIMO is first derived in theory, and then the user channel orthogonality is evaluated by the ratio of sum rate capacity in real channels to its upper bound. An empir-ical formula for the ratio is finally built based on field measurement in different setups.The formula could be employed to evaluate the impacts of different system parameters on channel capacity quantitatively, which is helpful in system design and evaluation.