Separability of Closely Spaced Users in Massive MIMO Systems

Massive multiple-input multiple-output (MIMO) evolved to a key enabling physical layer (PHY) technology for the fifth generation (5G) of mobile communication systems and beyond. While the envisioned use cases of such communications systems are diverse, so are the challenges to meet their respective requirements. As a large-scale evolution of already well-established MIMO communications technologies, massive MIMO promises benefits with respect to all possible use cases. Theoretical works on massive MIMO, however, typically assumes i.i.d. Rayleigh fading channels without spatial channel correlation. The application of this model is justified through the assumption of rich scattering environments, which is claimed to hold, for example, in indoor environments. Spatial correlation of wireless channels leads to inter-stream interference in single-user MIMO communications systems and to inter-user interference in multi-user MIMO systems with linear precoding. Channel correlation is therefore crucial for the performance of such MIMO techniques for mobile communications systems. I perform wireless channel measurements in a massive MIMO downlink setup for three different scenarios. By measuring the wireless channel at 148 user positions with half-wavelength spacing on a straight line, I investigate the spatial channel correlation with respect to inter-user distance. I perform a stationarity analysis of the measured wireless channel in the spatial domain in terms of receiver positions. The results for the spatial channel correlation at the receiver side and the achievable spectral efficiency (SE) show that the channel correlation is significant in a deep indoor scenario, even at large inter-user distances of many wavelengths. I propose a novel wireless channel model based on geometry, which allows the correct modeling of spatial channel correlation. By assigning explicit positions to antennas and scattering elements in a wireless communications scenario, I obtain a spatially consistent wireless MIMO channel model. Since the strength of a scattering event and the scattering element positions are input parameters of this model, the Rician K factor is adjustable. This property makes scenarios in between free space propagation and i.i.d. Rayleigh fading attainable. I provide a statistical analysis of this model to show that the amplitude distribution is Rician and provide an analytic approximation for the Rician K factor. I fit the measured massive MIMO downlink scenarios with the proposed spatially consistent wireless channel model. For this fit, I distribute the scattering elements spatially according to the respective measured scenario and adjust the scattering event strength to match the Rician K factor of the measurement. The channel model fit demonstrates the importance of scattering element placement and its impact on the spatial channel correlation. This shows that the Rician K factor is not sufficient to characterize spatial channel correlation. I perform an antenna array design to decrease the inter-user channel correlation of closely spaced users and thereby increase the user separability. First, I investigate the impact of antenna array aperture size on the achievable SE for three different linear antenna array configurations. Secondly, I perform antenna array design by array thinning to determine the individual antenna element positions within a linear array. These antenna array design methods yield increased performance in terms of achievable sum SE and improve reliability in terms of minimal achievable SE. Since the array design is nonadaptive and the number of antenna elements is fixed, these improvements do not require increased system complexity. In my work, I show the importance of spatial correlation of wireless channels for massive MIMO systems analytically, through measurements, and through simulation. Since the correct modeling of this correlation is necessary for accurate performance evaluation, I propose a spatially consistent wireless channel model for massive MIMO. I investigate the antenna array property of user separability and perform array design to reduce the spatial channel correlation.