Network MIMO large-system analysis and the impact of CSIT Estimation

We consider the downlink of a cellular network with multiple cells, multi-antenna base stations and single-antenna user terminals, including arbitrary inter-cell cooperation clusters, realistic distance-dependent pathloss and general “fairness” requirements. We focus on linear zero-forcing beamforming. Beyond Monte Carlo simulation, no efficient computation method to evaluate the ergodic throughput of such systems has been provided so far. We propose a method based on the combination of some large random matrix results with convex optimization. The proposed method is computationally much more efficient than Monte Carlo simulation and provides further insight in the joint operations of multiuser downlink beamforming, inter-cell cooperation and interference and opportunistic “fair” scheduling. Further, our analysis can be easily extended to consider achievable rates in the presence of non-ideal channel state information. In particular, we can study the system capacity of schemes based on explicit downlink training and channel state feedback.