How Does Performance Scale with Antenna Number for Extremely Large-Scale MIMO?

Abstract: Extremely large-scale multiple-input multiple-output (XL-MIMO) communications correspond to systems whose antenna size is so large that conventional assumptions, such as uniform plane wave (UPW) impingement, are no longer valid. This paper studies the channel modelling and performance analysis of XL-MIMO communication based on the generic spherical wavefront propagation model. First, for the single-user uplink/downlink communication with the optimal maximum ratio combining/transmission (MRC/MRT), we rigorously derive a new closed-form expression for the resulting signal-to-noise ratio (SNR), which includes the conventional SNR expression based on UPW assumption as a special case. Our result shows that instead of scaling linearly with the base station (BS) antenna number $M$, the SNR with the more generic spherical wavefront model increases with $M$ with diminishing return, governed by a new parameter called angular span. One important finding from our derivation is the necessity to introduce a new distance criterion, termed critical distance, to complement the classical Rayleigh distance for separating the near- and far-field propagation regions. While Rayleigh distance is based on the phase difference across array elements and hence depends on the electrical size of the antenna, the critical distance cares about the amplitude/power difference and only depends on its physical size. We then extend the study to the multi-user XL-MIMO communication system, for which we demonstrate that inter-user interference (IUI) can be mitigated not just by angle separation, but also by distance separation along the same direction. This offers one new degree of freedom (DoF) for interference suppression with XL-MIMO.