Full year channel characterization for adaptive underwater acoustic communications

Abstract

The spectral efficiency of current underwater acoustic communication links is less than 1 bit per second per Hz of bandwidth. For current 3G cellular networks it is around 3 b/s/Hz and for the next generation cellular systems it will be ~15 b/s/Hz. This discrepancy reflects the challenging nature of the ocean as a communication channel — poor transmission characteristics that are strongly frequency dependent, large signal spreading in both time and frequency, and long delay. As a first step to improving matters, in this paper we describe a methodology for more incisive channel characterization for adaptive underwater communications than has been routinely available in the past and illustrate its application to the area around National Data Buoy Center (NDBC) buoy 44014 (36.611N, 74.836W, water depth 47.5 m). The interaction of sound with the sea surface is important for underwater acoustic communications. In particular, at most temperate and higher latitude sites the extent to which underwater sound is shielded from the surface has a very strong seasonal dependence with the surface being least shielded when it is roughest. The moving rough sea surface causes an angular redistribution and Doppler shift of sound that interacts with it. The surface wave spectrum from 0.03 Hz to 0.5 Hz is routinely measured by wave buoys. The 15 cm and shorter acoustic waves of interest in underwater acoustic communications are, however, sensitive to frequencies in the surface wave spectrum well above 0.5 Hz. At those frequencies, measuring wave spectra is difficult. There have been attempts to fit the copious data at low frequencies and the sparse data at higher frequencies to produce analytical spectra depending on a few parameters that describe the spectrum from a few hundredths of a Hz up to 5 Hz and above. In our work, meteorological and oceanic data measured by NDBC 44014 and model estimates of downward radiation fluxes at its location are used to produce a time series, covering all of 2004 with 3 hr spacing, of the input parameters (friction velocity, inverse wave age, and wind speed) needed by the spectral model of Elfouhaily et al. [1]. For each of March, June, September, and December we find a time when the wind speed is near the 75th percentile wind speed for that month. We input the wind speed for that time and its associated friction velocity and inverse wave age to the spectral model to get a spectrum and then generate sea surface realisations of that spectrum. Using a rough surface parabolic equation model [2] we compute a channel impulse response for each surface. We use a deep source and show results for deep and mid-water column receivers.