In this paper， the upper limit of sampling rate of chaotic signals which are generated by an optoelectronic oscillator and used as the entropy source of physical random number is studied. The effects of different sampling rates on the random characteristics of the chaotic entropy source are analyzed. The maximum sampling rate for random sampling is obtained. The random characteristics of the signal are characterized by the autocorrelation function and power spectra. The maximum sampling rate is numerically studied by the run test and normality test. Simulation results show that， for the random sampling of chaotic signals， there exists an upper limit of sampling rate. On this basis， the maximum random sampling rate and the maximum normal sampling rate are obtained by calculations. The changes of these two maximum sampling rates with the feedback intensity of the optoelectronic oscillation system are numerically studied.