Favorable Propagation Studies for Massive MIMO Systems

Abstract

Massive MIMO (multiple-input multiple-output) is a key technology for 5G/6G networks. Its main advantages lies in enhanced spectral/energy efficiency and simplified processing in multi-user scenarios. These benefi ts are attributed to a phenomenon known as favorable propagation (FP). In this thesis, we study the asymptotic FP (as the number N of antennas increases without bound) for various scenarios and array con gurations. In particular, we establish the asymptotic FP for uniform circular/cylindrical arrays under fixed element spacing. To do so, a novel technique is developed based on a Bessel series expansion. The impact of grating lobes (GL) on the asymptotic FP of uniform linear/planar arrays is analyzed. A novel design of non-uniform linear arrays, based on subarray structure, is proposed to eliminate the impact of GL. This approach is robust in the frequency domain and is applicable to wideband systems. The impact of location and phase errors on favorable propagation is studied. It is shown that the asymptotic FP holds for perturbed arrays if and only if it holds for the unperturbed ones, for any i.i.d distribution of nite variance. While errors have negligible asymptotic effects, they significantly affect the rate of convergence with N: it slows down from 1=N2 (no errors) to 1=N (with errors), so that more antennas are needed in the later case to attain high SINR. Next, we consider nonasymptotic scenario and to reduce the array complexity, minimize the number of antennas subject to SINR constraints. Despite the non-convex nature of resulting optimization problems, globally optimal closed-form solutions are obtained. The number of antennas can be reduced by almost 50% if variable antenna spacing is allowed compared to the fixed spacing of half a wavelength.