Abstract: From the broadband data recorded by the digital seismic network of Jiangsu Province in 2011, this paper selected two station groups A and B (each has four stations). The two groups were under different ambient noise area and the average station distances were 44.6 km and 30.5 km, respectively. Firstly, by the sliding window cross-correlation technique, the offsets of the Green’s function were measured by ambient noise cross correlation between station-pairs in the year 2011. Then, the singular value decomposition was applied to solve the over-determined equation of ΔD_Si-Sj and the coefficient matrices, so as to calculate clock error ΔS_i for a single station. And the covariance matrix was introduced to estimate the error range of calculation results under different confidence level. The results showed that ignoring the obvious clock error ( >3 s), the average RMS of time errors for eight stations was about 0.4215 s, and on the 95% confidence level, the errors ranges for the two groups of stations were ±0.454 4s and ±0.4283, respectively. The average travel time residual was about 0.386 s by adopting HYPOSAT to locate actual earthquakes in Jiangsu region during the period of 2010—2011. Therefore, both residuals are consistent in accuracy, suggesting it is reliable to estimate the time error for a single station based on Green’s function by noise cross-correlation.