The principle of phase diversity(PD) is firstly introduced. The effect of the benchmark sub-mirror of sparse aperture imaging system on PD is illustrated. Taking Golay3 sparse aperture as an example, the object is imaged by each sub-mirror of the imaging system. The arithmetic mean standard deviation(AMSD) between images of each sub-mirror and the original object is calculated, and the sub-mirror with minimum AMSD value is selected as the benchmark sub-mirror. The influence of errors of the benchmark sub-mirror which are unknown practically on the accuracy of error estimation of the other sub-mirrors is analyzed. The impact of the defocus amount on the results of PD is also discussed. It is shown that the benchmark sub-mirror must be confirmed before error estimation of the other sub-mirrors of the sparse aperture imaging system with PD. The smaller the AMSD value is, the smaller the error of the corresponding sub-mirror is, the smaller the error of the residual sub-mirrors in the system calculated by PD method is, while the variation of the defocus amount has little influence on PD results.