Abstract
Training-based transmission over Rayleigh block-fading multiple-input multiple-output (MIMO) channels is investigated. As a training method a combination of a pilot-assisted scheme and a biased signaling scheme is considered. The achievable rates of successive decoding (SD) receivers based on the linear minimum mean-squared error (LMMSE) channel estimation are analyzed in the large-system limit, by using the replica method under the assumption of replica symmetry. It is shown that negligible pilot information is best in terms of the achievable rates of the SD receivers in the large-system limit. The obtained analytical formulas of the achievable rates can improve the existing lower bound on the capacity of the MIMO channel with no channel state information (CSI), derived by Hassibi and Hochwald, for all SNRs. The comparison between the obtained bound and a high-SNR approximation of the channel capacity, derived by Zheng and Tse, implies that the high-SNR approximation is unreliable unless quite high SNR is considered. Energy efficiency in the low-SNR regime is also investigated in terms of the power per information bit required for reliable communication. The required minimum power is shown to be achieved at a positive rate for the SD receiver with no CSI, whereas it is achieved in the zero-rate limit for the case of perfect CSI available at the receiver. Moreover, numerical simulations imply that the presented large-system analysis can provide a good approximation for not so large systems. The results in this paper imply that SD schemes can provide a significant performance gain in the low-to-moderate SNR regimes, compared to conventional receivers based on one-shot channel estimation.
Introduction
In Multiple-Input Multiple-Output (MIMO) communications, the availability of Channel State Information (CSI) at the receiver is often a prerequisite for high-performance decoding. However, in practical scenarios—especially those with short coherence times—the receiver must estimate the channel using pilot symbols. This process consumes valuable bandwidth and power resources. Traditional "one-shot" estimation methods, such as those analyzed in the well-known Hassibi-Hochwald (HH) bound, often underestimate the achievable rate because they do not account for the potential to refine channel estimates using decoded data symbols as "virtual pilots."
Successive Decoding and Biased Signaling
To address these limitations, this work investigates a Successive Decoding (SD) receiver architecture. In this scheme, transmission is organized into stages and substages to facilitate iterative refinement of the channel estimate.
As illustrated in Figure 1, the decoding process proceeds sequentially. Symbols decoded in earlier stages are utilized by a Linear Minimum Mean-Squared Error (LMMSE) estimator to improve channel knowledge for subsequent stages. To initiate this process without heavy pilot overhead, the authors employ a biased signaling scheme. In this approach, data symbols possess a known probabilistic bias, providing enough initial information for the receiver to begin the iterative estimation and decoding cycle even with minimal dedicated pilots.
Large-System Analysis via the Replica Method
The analytical framework leverages the "replica method" from statistical physics to evaluate performance in the large-system limit. By assuming the number of antennas ($M, N$) and the coherence time ($T_c$) grow to infinity while maintaining constant ratios, the complex random matrix problem is reduced to a set of deterministic fixed-point equations. This approach, under the assumption of replica symmetry, allows for the derivation of closed-form expressions for achievable rates that remain accurate for finite systems.
Key Findings and Results
The analysis yields several critical insights into the performance of blind MIMO receivers:
- Pilot Minimalism: In the large-system limit, the optimal amount of dedicated pilot information is found to be negligible. The SD receiver effectively learns the channel from the data itself through successive feedback.
- Tighter Bounds: The derived formulas provide a lower bound on the capacity that is strictly tighter than the HH bound across all signal-to-noise ratios (SNRs), typically offering a gain of 1 to 1.8 dB.
- Energy Efficiency Shift: Unlike the perfect CSI case where energy efficiency is maximized as the communication rate approaches zero, the "No CSI" SD receiver achieves minimum power per bit at a positive rate. This highlights the necessity of a minimum data flow to facilitate effective channel learning.
- High-SNR Discrepancy: The results indicate that the Zheng-Tse high-SNR approximation is overly optimistic and only converges with actual achievable rates at extremely high SNR levels, making it less reliable for practical system design at moderate SNRs.
Conclusion
This study demonstrates that successive decoding schemes can provide a significant performance gain in the low-to-moderate SNR regimes compared to conventional receivers based on one-shot channel estimation. While the analysis relies on the replica symmetry assumption and i.i.d. Rayleigh fading, numerical simulations confirm its accuracy even for relatively small systems (e.g., 8x8 configurations). Future research may extend this framework to spatially correlated MIMO channels, where antenna proximity introduces new trade-offs between channel estimation and diversity.
Summary updated