Theoretical Performance Limits of Massive MIMO With Uncorrelated Rician Fading Channels

This paper considers a Massive MIMO network with <inline-formula> <tex-math notation="LaTeX">$ {L}$ </tex-math></inline-formula> cells, each comprising a base stations (BS) with <inline-formula> <tex-math notation="LaTeX">$ {M}$ </tex-math></inline-formula> antennas and <inline-formula> <tex-math notation="LaTeX">$K$ </tex-math></inline-formula> single-antenna user equipments. Within this setting, we are interested in deriving approximations of the achievable rates in the uplink and downlink under the assumption that single-cell linear processing is used at each BS and that each intracell link forms an uncorrelated MIMO Rician fading channel matrix; that is, with a deterministic line-of-sight (LoS) path and a stochastic non-LoS component describing a spatial uncorrelated multipath environment. The analysis is conducted assuming that <inline-formula> <tex-math notation="LaTeX">$ {N}$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$ {K}$ </tex-math></inline-formula> grow large with a given ratio <inline-formula> <tex-math notation="LaTeX">${\textbf {N/K}}$ </tex-math></inline-formula> under the assumption that the data transmission in each cell is affected by channel estimation errors, pilot contamination, an arbitrary large scale attenuation and LoS components. Numerical results are used to prove that the approximations are asymptotically tight, but accurate for systems with finite dimensions. The asymptotic results are also used to evaluate the impact of LoS components. In particular, we exemplify how the number of antennas for achieving a target rate can be substantially reduced with LoS links of only a few dBs of strength.