Performance study of variable-rate modulation for underwater communications based on experimental data

In this paper, we present an analysis of the performance of a variable-rate adaptive modulation system based on instantaneous SNR information. The SNR traces we consider are part of the SubNet'09 experimental dataset, and have been derived by processing an hyperbolic frequency-modulated signal in the 9-14 kHz band. We start by deriving the high level behavior of the channel in terms of the statistics of channel fading effects, which are found to be well modeled by a Nakagami-m distribution, where the parameter m is estimated over a whole experiment or over smaller time windows throughout the experiment, depending on the variability of the SNR. The statistics of the channel behavior are then used to derive the performance of a variable-rate modulation scheme switching between five different constellations; the cases of both instantaneous and outdated channel knowledge are considered. Analytical results are compared to simulations to show that the Nakagami-m distribution can satisfactorily capture the statistics of the channel, provided that the estimation of the m parameter, as well as of the correlation of the SNR process, is repeated in case of macroscopic channel variations.