The influence of the diffusion random noise on the applicability of the integration method to a bimodal distribution is systematically studied by computer simulation. A two-dimensional parameter space is constructed to represent all bimodal distributions. Four error functions are then defined in this parameter space to evaluate the integration method. It is found: (i) for any noise level, there exists a good area in parameter space where the integration method works successfully; (ii) as the noise level increases, the good area shrinks to a region where only those slow modes, having larger weightings and seperated far from the fast mode, could be well resolved; (iii) it is mainly the noise level, not the diffusion cosine period, which limits the applicability of the integration method. Our numerical study provides useful information for a quasi-elastic light scattering experiment to decide whether the integration method is applicable.
