Distributed Massive MIMO in Millimetre Wave Communication

This thesis considers a distributed massive MIMO (D-MaMIMO) system for millimetre wave (mmWave) communication for an outdoor coverage area, as the wavelength of mmWave makes it easier and the free space path loss necessitates the use of relatively large distributed antenna arrays. We assume that the line-of-sight (LoS) links are available between the access points (APs) and user equipment (UEs). We have examined different AP arrangements to serve a given square area using the beamforming (BF) for a single user case. Furthermore, the zero-forcing (ZF) pre-coding is applied at a central processing unit (CPU) on the downlink to separate multiple users. We focus on these multi-user scenarios with varying numbers of APs to demonstrate the extent to which closely spaced users can be separated by ZF processing. We examine the determinant of the effective composite channel matrix to demonstrate the conditions under which the ZF problem may become ill-conditioned. We then show that nearly perfect separation is attainable, even when the UEs are only a few metres apart. Subsequently, an eigenvalue decomposition (EVD) based ZF is proposed to improve the performance of multi-antenna UEs. It has been observed that 3DBF has limited scope in circumstances when users are distributed horizontally, near to the same height as the APs and it is advantageous to employ non-square AP antenna arrays to maximize azimuth separation, especially for multi-user environments. The throughput per UE indicates how many users could be served effectively using the aforementioned schemes and AP arrangements for these multi-user cases. We further explore the significant issue of multipath propagation characteristics for mmWave communication and propose the novel distinction between the effective and the environmental K-factor for Ricean channels. A closed-form approximation for the effective K-factor is derived and corroborated by comparison with numerical results.