Optimal Throughput-Diversity-Delay Tradeoff in MIMO ARQ Block-Fading Channels

TL;DR

This paper analyzes the optimal tradeoff between throughput, diversity, and delay in MIMO ARQ channels with practical finite constellations, providing theoretical bounds and demonstrating near-optimal coding performance.

Contribution

It introduces a comprehensive analysis of the throughput-diversity-delay tradeoff in MIMO ARQ channels with finite constellations, extending previous work to more practical scenarios.

Findings

Optimal SNR exponent matches a modified Singleton bound.

Practical MDS codes achieve the optimal SNR exponent.

Analysis applies to both short-term and long-term static fading cases.

Abstract

In this paper, we consider an automatic-repeat-request (ARQ) retransmission protocol signaling over a block-fading multiple-input, multiple-output (MIMO) channel. Unlike previous work, we allow for multiple fading blocks within each transmission (ARQ round), and we constrain the transmitter to fixed rate codes constructed over complex signal constellations. In particular, we examine the general case of average input-power-constrained constellations as well as the practically important case of finite discrete constellations. This scenario is a suitable model for practical wireless communications systems employing orthogonal frequency division multiplexing techniques over a MIMO ARQ channel. Two cases of fading dynamics are considered, namely short-term static fading where channel fading gains change randomly for each ARQ round, and long-term static fading where channel fading gains remain constant over all ARQ rounds pertaining to a given message. As our main result, we prove that for the block-fading MIMO ARQ channel with discrete input signal constellation satisfying a short-term power constraint, the optimal signal-to-noise ratio (SNR) exponent is given by a modified Singleton bound, relating all the system parameters. To demonstrate the practical significance of the theoretical analysis, we present numerical results showing that practical Singleton-bound-achieving maximum distance separable codes achieve the optimal SNR exponent.