Large-System Analysis of Correlated MIMO Multiple Access Channels with Arbitrary Signaling in the Presence of Interference

TL;DR

This paper analyzes large correlated MIMO multiple-access channels with interference, deriving ergodic sum-rate expressions for arbitrary signaling, and explores precoding strategies to enhance performance in interference-limited scenarios.

Contribution

It provides the first large-system analysis of correlated MIMO multiple-access channels with interference using the replica method, including arbitrary signaling and precoding optimization.

Findings

Interference with discrete constellations is easier to manage than Gaussian interference.

Optimized precoding significantly improves achievable rates at low-to-mid SNRs.

Asymptotic sum-rate expressions enable better system design in large MIMO networks.

Abstract

Presence of multiple antennas on both sides of a communication channel promises significant improvements in system throughput and power efficiency. In effect, a new class of large multiple-input multiple-output (MIMO) communication systems has recently emerged and attracted both scientific and industrial attention. To analyze these systems in realistic scenarios, one has to include such aspects as co-channel interference, multiple access and spatial correlation. In this paper, we study the properties of correlated MIMO multiple-access channels in the presence of external interference. Using the replica method from statistical physics, we derive the ergodic sum-rate of the communication for arbitrary signal constellations when the numbers of antennas at both ends of the channel grow large. Based on these asymptotic expressions, we also address the problem of sum-rate maximization using statistical channel information and linear precoding. The numerical results demonstrate that when the interfering terminals use discrete constellations, the resulting interference becomes easier to handle compared to Gaussian signals. Thus, it may be possible to accommodate more interfering transmitter-receiver pairs within the same area as compared to the case of Gaussian signals. In addition, we demonstrate numerically for the Gaussian and QPSK signaling schemes that it is possible to design precoder matrices that significantly improve the achievable rates at low-to-mid range of signal-to-noise ratios when compared to isotropic precoding.