A model called SYTRAC.L, for the prediction of echoes from large, hard reflectors, facing finite dimensions transducers, is proposed. For great reflectors, the introduction of the acoustic ray theory allows low cpu costs and more reasonable memory needs, with comparison with other methods where a numerical integration is performed on the reflector, since this integration is eliminated. In a first stage, this model does not take into account reflector edge diffraction effects, but these are being introduced with GTD approaches. The validation of the formulation and of the numerical implementation was done by comparison with analytical results in two canonical configurations. Experimental comparisons for plane and cylindrical reflectors have also been performed. This model is used to estimate the accuracy of telemetries on large structures, at distances less or greater than the far-field limit. As an example, results of a parametric study of echo characteristics from cylindrical and plane targets are described according to their distance and tilting.