A precise comparison of the surface and volume approaches of the Ffowcs Williams acoustic integral formulation is conducted for a propeller in transonic operating conditions. For both approaches, the calculations are carried out directly starting from CFD input data provided in the propeller rotating frame, i.e. with supersonically moving emission points. The principle and the calculation algorithm on which this particular integration technique is based on are reminded. Then calculations carried out for four CFD meshes of different densities show that the volume method is slightly less sensitive to the numerical dissipation of the aerodynamic computations than the surface method. These calculations also show that, to be of more interest, the volume method requires a specific way of meshing the flow. In addition, two techniques for determining the regions of the dominant acoustic sources are explored. With the first one, a rather conventional technique based on the local quadrupole term, the results show that specific terms, chosen according to the concerned phenomenon, may be better indicators of the real noise sources than the original shear and entropy terms. The second one, less known and consisting in calculating the elementary acoustic time signature radiated by each cell of the grid, seems more effective but may turn out to be costly in terms of data storage with the volume method.