On the Capacity of Large-MIMO Block-Fading Channels

Wei Yang, Giuseppe Durisi, Erwin Riegler

TL;DR

This paper proves that a common technique for maximizing data rates in wireless communication systems (Unitary Space-Time Modulation or USTM) is not optimal for large MIMO systems when the channel changes rapidly. They introduce a new, more effective modulation scheme called Beta-variate Space-Time Modulation (BSTM) that can achieve up to a 13% rate gain in practical scenarios.

What Problem Does This Paper Address?

Wireless communication systems often use multiple antennas (MIMO) to boost data rates. The theoretical maximum data rate, or "capacity," of these systems depends heavily on whether the transmitter and receiver know the exact characteristics of the wireless channel (which constantly changes due to fading).

In the "coherent" setting, where channel state information is known, capacity is well-understood. However, in the more realistic "noncoherent" setting, where this information is unknown, calculating capacity is much harder. A key challenge arises in "block-fading" channels, where the channel remains constant for a short block of time (the "coherence time") and then changes independently.

For a long time, a technique called Unitary Space-Time Modulation (USTM) was believed to be capacity-achieving in the high signal-to-noise ratio (SNR) regime for these noncoherent MIMO channels. However, this was only proven for cases where the channel coherence time was sufficiently long relative to the number of antennas. A conjecture by Zheng & Tse (2002) suggested USTM might not be optimal for "large-MIMO" systems, where the number of antennas is very large, especially when the coherence time is short. This paper tackles this conjecture, aiming to characterize the true capacity and find the optimal transmission strategy in this challenging large-MIMO, short-coherence-time regime.

Methods

The authors' core contribution lies in proving that USTM is not capacity-achieving in the large-MIMO, short-coherence-time regime. They achieve this by:

  • Duality Approach: They leverage a powerful mathematical tool called duality, which allows them to derive tight upper bounds on the channel capacity by carefully choosing a probability distribution for the received signal.
  • Novel Input Distribution (BSTM): They introduce and analyze a new input signal distribution, termed Beta-variate Space-Time Modulation (BSTM). This distribution is constructed from a product of a unitary random matrix and a diagonal matrix whose entries are derived from the square-root of eigenvalues of a Beta-distributed random matrix. This BSTM distribution is shown to be capacity-achieving in the high-SNR regime for the large-MIMO, short-coherence-time scenario.
  • Geometric Structure Exploitation: Their proof technique utilizes the geometric structure of the MIMO block-fading channel's input-output relationship, building upon prior work but employing a different set of mathematical tools.
  • Probability Density Function (PDF) Characterization: They derive a novel closed-form characterization for the PDF of the MIMO block-fading channel output, which generalizes previous results and simplifies the capacity derivation.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • USTM is Not Always Optimal: The most significant insight is the definitive proof that USTM, while optimal in some scenarios, fails to achieve capacity in large-MIMO systems when the channel coherence time is short relative to the number of antennas. This settles a long-standing conjecture.
  • The Importance of Input Distribution: The paper highlights that the specific distribution of the transmitted signal (the input distribution) is critical for achieving capacity, especially in complex fading environments.
  • Beta-Variate Distribution's Role: The introduction of BSTM, derived from Beta-distributed random matrices, is a key innovation. This distribution effectively captures the statistical properties of the channel in the large-MIMO, short-coherence-time regime, leading to improved performance.
  • Rate Gain in Large Systems: The findings are particularly relevant for "large-MIMO" systems, where the number of antennas is substantial. In these scenarios, the benefits of using BSTM over USTM become more pronounced.

Key Results

  • Capacity Characterization: The paper provides a high-SNR characterization of the capacity for Rayleigh block-fading MIMO channels in the large-MIMO setting (T < M + N), generalizing existing results.
  • BSTM Achieves Capacity: They demonstrate that BSTM is the capacity-achieving input distribution in this regime.
  • Rate Gain: Numerical results show that using BSTM instead of USTM can yield a rate gain of up to 13% for practical SNR values (e.g., 30 dB) when the number of receive antennas (N) is significantly larger than the channel coherence time (T) and M = min{[T/2], N}. This gain grows logarithmically with the number of receive antennas.

Limitations and Future Work

  • High-SNR Focus: The primary capacity characterization is for the high-SNR regime. While the authors suggest their results are accurate at moderate SNRs, a complete characterization across all SNR regimes for the large-MIMO case remains an area for further exploration.
  • Block-Fading Assumption: The analysis is based on the block-fading model. Future work could investigate the impact of more complex fading models (e.g., frequency-selective fading) on the capacity and optimal transmission strategies.
  • Practical Implementation: While the paper provides theoretical insights, the practical implementation of BSTM, including its complexity and robustness, would be a valuable next step.
  • Generalization to Other Distributions: The paper focuses on Rayleigh fading. Extending the analysis to other fading distributions could broaden the applicability of the findings.

Summary updated

Abstract

We characterize the capacity of Rayleigh block-fading multiple-input multiple-output (MIMO) channels in the noncoherent setting where transmitter and receiver have no a priori knowledge of the realizations of the fading channel. We prove that unitary space-time modulation (USTM) is not capacity-achieving in the high signal-to-noise ratio (SNR) regime when the total number of antennas exceeds the coherence time of the fading channel (expressed in multiples of the symbol duration), a situation that is relevant for MIMO systems with large antenna arrays (large-MIMO systems). This result settles a conjecture by Zheng & Tse (2002) in the affirmative. The capacity-achieving input signal, which we refer to as Beta-variate space-time modulation (BSTM), turns out to be the product of a unitary isotropically distributed random matrix, and a diagonal matrix whose nonzero entries are distributed as the square-root of the eigenvalues of a Beta-distributed random matrix of appropriate size. Numerical results illustrate that using BSTM instead of USTM in large-MIMO systems yields a rate gain as large as 13% for SNR values of practical interest.
BibTeX
@article{yang2013capacity,
  author = {Wei Yang and Giuseppe Durisi and Erwin Riegler},
  title = {{On the Capacity of Large-MIMO Block-Fading Channels}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2013},
  url = {https://openreview.net/forum?id=nA0pXisWhQ},
}
Figure