Fluorescence correlation spectroscopy (FCS) provides most commonly values of diffusion coefficients of fluorescently labeled species in a solution, especially biopolymers. A newly developed procedure for the determination of diffusion coefficient distributions applicable to polydisperse polymers or nanoparticles, based on the well-known CONTIN algorithm, is described and tested on both simulated FCS correlation functions and real experimental data. Good resolution of bimodal distributions is observed, and it is quantitatively established how the resolution depends on the level of experimental noise. Effects of incorrect calibration of the focal volume on the obtained diffusion coefficient and its distribution are described for single-focus FCS. With rapid expansion of the FCS method from biology and biochemistry into the polymer field, and recently into many other disciplines including, e.g., environmental sciences, such a technique was very much needed.