Diversity versus Multiplexing at Finite Blocklength

Johan Östman, Wei Yang, Giuseppe Durisi, Tobias Koch

Abstract: A finite blocklenth analysis of the diversity-multiplexing tradeoff is presented, based on nonasymptotic bounds on the maximum channel coding rate of multiple-antenna block-memoryless Rayleigh-fading channels. The bounds in this paper allow one to numerically assess for which packet size, number of antennas, and degree of channel selectivity, diversity-exploiting schemes are close to optimal, and when instead the available spatial degrees of freedom should be used to provide spatial multiplexing. This finite blocklength view on the diversity-multiplexing tradeoff provides insights on the design of delay-sensitive ultra-reliable communication links.

Introduction

Traditional wireless communication theory often relies on the Diversity-Multiplexing Tradeoff (DMT) analyzed under the assumption of infinite blocklengths and high signal-to-noise ratios (SNR). However, the rise of Ultra-Reliable Low-Latency Communication (URLLC) in 5G systems introduces a new paradigm: sending very short packets (e.g., 100-200 symbols) with extremely low error tolerances ($10^{-3}$ to $10^{-5}$). In this "finite blocklength" regime, classical asymptotic formulas break down because they neglect the significant overhead required to estimate channel conditions within a narrow temporal or spectral window.

Methodology

This study employs tools from finite blocklength information theory to derive tight non-asymptotic bounds on the maximum achievable coding rate. The analysis focuses on a Rayleigh block-fading model where the channel remains constant for a coherence interval ($T$) before changing. Crucially, the model assumes a non-coherent setting where neither the transmitter nor the receiver has a priori channel state information (CSI), reflecting the realities of high-mobility or low-latency scenarios.

Performance is evaluated using:

  • Dependency Testing (DT) lower bound: To demonstrate achievable rates.
  • Meta-converse (MC) upper bound: To establish the absolute physical limits.
  • Unitary Space-Time Modulation (USTM): A multiplexing-oriented scheme known for efficiency in non-coherent channels.

Numerical evaluations are performed using parameters consistent with the 3GPP LTE standard, such as a packet length ($n$) of 168 symbols and error probabilities ($\epsilon$) ranging from $10^{-3}$ to $10^{-5}$.

Key Insights: The Cost of Learning

A central finding is that the achievable rate is not monotonic with respect to the channel's coherence time. There exists a "sweet spot" ($T^*$) for performance. If the channel changes too rapidly (high selectivity), too much resource is wasted on channel estimation. If it stays constant for too long, the system loses the benefits of time-frequency diversity.

Coding rate bounds for a 2x2 MIMO system with n=168 and error probability 10^-3.

As illustrated in Figure 1, for a $2 \times 2$ MIMO system, diversity-only schemes like the Alamouti code are near-optimal only when the channel provides minimal diversity branches ($L \approx 1$). As the number of diversity branches increases, multiplexing-oriented schemes like USTM quickly become superior. For instance, at $L=14$, the gap between USTM and Alamouti is approximately 0.5 bits/channel use. Furthermore, the standard "outage capacity" metric ($C_{\text{out}, \epsilon}$) is shown to be overly optimistic as it ignores the channel estimation overhead.

Scaling with Antenna Configurations

The advantage of multiplexing-oriented designs becomes even more pronounced as the number of antennas increases.

Coding rate bounds for a 4x4 MIMO system with n=168 and error probability 10^-3.

In the $4 \times 4$ configuration (Figure 2), the performance gap between DT-USTM and diversity-exploiting schemes like Frequency-Switched Transmit Diversity (FSTD) is substantial. While tightening reliability requirements (e.g., to $10^{-5}$) makes diversity schemes more competitive, multiplexing remains the dominant strategy as channel selectivity increases.

Conclusion

This research demonstrates that in the finite blocklength regime, the "cost of learning" the channel significantly shifts the optimal balance between diversity and multiplexing. For URLLC applications, designers must move beyond traditional outage capacity metrics and consider non-asymptotic bounds to accurately assess the performance of short-packet MIMO systems. Diversity-exploiting schemes are only optimal in very specific, low-selectivity conditions; in most practical URLLC scenarios, spatial multiplexing provides superior throughput.

Summary updated

Abstract

A finite blocklenth analysis of the diversity-multiplexing tradeoff is presented, based on nonasymptotic bounds on the maximum channel coding rate of multiple-antenna block-memoryless Rayleigh-fading channels.The bounds in this paper allow one to numerically assess for which packet size, number of antennas, and degree of channel selectivity, diversity-exploiting schemes are close to optimal, and when instead the available spatial degrees of freedom should be used to provide spatial multiplexing. This finite blocklength view on the diversity-multiplexing tradeoff provides insights on the design of delay-sensitive ultra-reliable communication links.
BibTeX
@inproceedings{ostman2014diversity,
  author = {Johan Östman and Wei Yang and Giuseppe Durisi and Tobias Koch},
  title = {{Diversity versus Multiplexing at Finite Blocklength}},
  booktitle = {ISWCS},
  year = {2014},
  url = {https://openreview.net/forum?id=-T2acLk4Ay5},
}
Figure