On the Behaviour of MIMO Systems with Ray-Based Models

Multiple-input-multiple-output (MIMO) is a key enabling technology to provide orders of magnitude greater data rates than single-antenna wireless communication systems. Channel observations from multiple antennas allow a MIMO base station (BS) to spatially separate users using simple digital linear processing techniques, serving them simultaneously on the same carrier frequency. The simplest processor achieves optimal performance under favourable propagation (FP), describing the theoretical condition in which the channels from two arbitrary users become mutually orthogonal when observed by an infinite number of BS antennas. Existing literature provides ample mathematical analysis of MIMO systems using classical statistical channel models due to their mathematical tractability. These are not physically-based, and are constructed using several assumptions about the propagation statistics, producing results which elude convincing physical interpretation as key features such as the array structure and ray statistics are not modelled. Few studies have analysed ray-based channels, which model each channel as a combination of scattered rays and form the basis of several standardised channel models including the 3GPP and COST models. The ray-based structure creates channels with embedded correlation and statistics which are directly dependent on the physically measurable characteristics, such as the ray statistics or antenna topology. Analytical results using ray-based models with generic parameterisation can represent a variety of propagation scenarios, and facilitate intuitive understanding of the effects of several physical properties on MIMO behaviour. However, existing studies which provide analytical results for ray-based models assume a uniform ray angular distribution, which simplifies analysis, but is rarely observed in practice. Additionally, our results show that the uniform angular distribution provides optimistic performance and obscures the sensitivity to antenna topology which is observed under more realistic angular distributions. This thesis investigates MIMO behaviour using a generic ray-based model. We provide analytical expressions for metrics pertaining to the performance with linear processors, antenna correlation, convergence to FP, and distance from FP for a finite-antenna system. These expressions indicate that these fundamental quantities are all intrinsically linked to a single variable, κ, describing the average interaction of rays. Higher values of κ caused by stronger average ray interference indicate greater inter-user channel interference and antenna correlation. The performance with linear processors is also mathematically shown to decrease with higher values of κ, confirming the existing understanding that MIMO thrives with diverse propagation. Examination of these results for a range of model parameterisations yields insight into the effects of several physical qualities on MIMO performance. We conclude that a narrower spread of ray angles or smaller antenna spacing at the BS increases the average interference power of two rays, inflating the correlation and inter-user interference, and harming performance. A comparison of different antenna topologies reveals that larger azimuthal apertures facilitate lower average ray interactions, however under space constraints, topologies with greater elevation resolution are advantageous. Finally, the ray-based model is used to examine a MIMO system with a reconfigurable intelligent surface (RIS). The physically-based correlation embedded in the ray-based channels facilitates the design of interpolation techniques to create an efficient channel estimation scheme for RIS-assisted MIMO.