In this work, we have investigated the estimation of particles size, to prove the feasibility of a simplified acoustic system for continuous characterization of turbid water during underwater cave exploration with autonomous vehicles. Because of the nature and concentration level of sediments in natural water networks, we have identified the most suitable models for suspended sediments. They were proposed by Urick [1] and Sheng & Hay [2] assuming that particles are movable, rigid and at low concentration. These models have the advantage to provide a simple analytical expression with a limited number of input parameters. In order to test a very simplified prediction tool, these models are used to calculate attenuation only for monodisperse particle size distributions Experimentally, we used clay powder sifted to 40 μm, to create turbid water similar to cave water. This mixture characterised using laser diffraction spectroscopy shows a large and non-uniform particle distribution. From the measurement of the echo amplitude in reflexion mode, we deduce the attenuation coefficient in a frequency range from 1.5 MHz to 16 MHz. We used several transducers for different travel distances (3.5 cm to 7.5 cm) and different clay concentrations in volume (0.3 to 1.7 %). We obtained a linear dependence between attenuation coefficient and concentration for any tested frequency in total agreement with the selected model. Fitting experimental attenuation data with this model, the best agreement was obtained with a particle diameter of 70 μm. This value is in agreement with the standard Dv90 parameter provided by laser diffraction analysis. This is an encouraging result to validate our minimalist experimental setup and the selected model for a simplified implementable system for in situ characterization of turbid water.