Abstract
In MIMO Rayleigh channels, transmitter correlations are known to degrade the performance of space-time codes (STCs) in high SNR regimes. This paper demonstrates that in single-keyhole channels, the impact of spatial correlation is more nuanced. For general STCs, performance can actually improve when the number of transmit antennas ($M$) exceeds the number of receive antennas ($N$), depending on the interaction between the correlation matrix and the codeword difference matrix.
Introduction
Multiple-Input Multiple-Output (MIMO) systems are a cornerstone of modern wireless communication, leveraging multiple antennas to enhance reliability and throughput. A common assumption in MIMO research is the Rayleigh fading model, which assumes a rich scattering environment. In such environments, spatial correlation between antennas—often caused by limited physical spacing—is universally detrimental to performance at high signal-to-noise ratios (SNR).
However, real-world propagation often encounters "keyhole" or "pinhole" effects, where signals are forced through a narrow physical bottleneck, such as a long corridor or a small aperture. In these single-keyhole channels, the channel matrix's rank is restricted to one, regardless of the number of antennas. While previous studies focused on orthogonal STCs (OSTBCs) in these environments, this work investigates general, non-orthogonal STCs.
The Impact of Antenna Configuration
The core finding of this research is that the effect of correlation in keyhole channels is governed by the ratio of transmit ($M$) to receive ($N$) antennas.
As summarized in the table above, the behavior in single-keyhole channels diverges from Rayleigh channels specifically for non-orthogonal STCs when $M > N$. In this regime, the correlation matrix $\mathbf{P}$ can act as a form of beamforming, shaping the codeword difference matrix $\mathbf{\Delta}$ to make signals more distinguishable at the receiver. Conversely, if $M \le N$, correlation remains a performance-degrading factor, similar to the Rayleigh case.
Analytical Insights and Theorems
The authors utilize Pair-wise Error Probability (PEP) analysis in the asymptotic high SNR regime to derive two fundamental theorems:
- Theorem 1 ($M > N$): There always exists a correlation matrix that improves PEP performance, provided a specific eigenvalue condition of the code is met.
- Theorem 2 ($M \le N$): Spatial correlation always degrades performance, mirroring standard Rayleigh channel behavior.
This distinction explains why earlier studies on OSTBCs always found correlation to be harmful; orthogonal codes are mathematically constrained in a way that prevents them from benefiting from the "shaping" effect of correlation.
Empirical Validation
To verify these theoretical results, the authors conducted Monte Carlo simulations. The results for the $M > N$ case demonstrate that a carefully chosen correlation matrix can significantly outperform independent antenna configurations.
In Figure 1, the red curve (Correlation $P_1$) shows a clear improvement over the blue curve (Independent Transmission Antennas), providing approximately a 1.5 dB gain in the high SNR regime. This confirms that spatial correlation, typically viewed as a hindrance, can be leveraged as a design tool in specific physical environments.
Conclusion
This work provides a new perspective on spatial correlation in MIMO systems. By moving beyond the Rayleigh fading model to consider single-keyhole channels, it reveals that correlation is not inherently negative. For systems with more transmit than receive antennas using non-orthogonal codes, spatial correlation can be exploited to enhance performance, offering a new dimension for optimizing wireless communication in constrained environments.
Summary updated