This paper deals with the problem of the effect of discretization level and certain other parameters characterizing the measurement setup on accuracy of the process of determination of the sound radiation efficiency by means of the Discrete Calculation Method (DCM) described by Hashimoto (2001). The idea behind DCM consists in virtual division of an examined sound radiating structure into rectangular elements each of which is further assumed to contribute to the total radiation effect in the same way as a rigid circular piston having the surface area equal to this of the corresponding virtual element and vibrating in an infinite rigid baffle. The advantage of the method over conventional sound radiation efficiency measurement techniques consists in the fact that instead of acoustic pressure values, source (plate) vibration velocity amplitude values are measured in a selected number of regularly distributed points. In many cases, this allows to determine the sound radiation efficiency with sufficient accuracy, especially for the low frequency regime. The key part of the paper is an analysis of the effect of discretization level (i.e. the choice of the number of points at which vibration amplitude measurements are to be taken with the use of accelerometers) on results obtained with the use of the method and their accuracy. The problem of determining an optimum level of discretization for given excitation frequency range is a very important issue as the labor intensity (time-consuming aspect) of the method is one of its main flaws. As far as the technical aspect of the method is concerned, two different geometrical configurations of the measurement setup were tested.