In this study the general undersea acoustic channel is modelled as a random, time-varying, linear filter. A detailed description of the pertinent channel-signal characteristics is given and the ideas of quasistationarity and channel ambiguity function are introduced and explored.
An attempt is then made to relate in a meaningful way the so called "microscopic" and "macroscopic" approaches, in the case of a channel consisting of a set of randomly moving scatterers (Reverberation) and a channel consisting of a set of correlated scatterers, distributed over a time-varying, random surface (Surface Scattering).
The problem of near optimum detection and signal design, for communication over purely random, general (not necessarily Gaussian), linear, time-varying, very noisy undersea acoustic channels is then formulated and the results are specialized to the cases of the Reverberation and the Very Rough Surface Scattering channels.
A suboptimum approach is finally considered, for detection and signal design over purely random, general, linear, time- varying, undersea acoustic channels, under arbitrary signal to noise ratio conditions. The results of this analysis are specialized to the case of a channel with Gaussian statistics and additive, white, Gaussian noise at its output and related to the findings of the small signal to noise ratio case. The theory is then extended to include cases of large signal to noise ratios and additive, Gaussian noise, at the channel output, which has both colored and white components.