Abstract: In this paper, both analytic and numerical simulation methods were used to discuss the focal depth error resulted from the local seismic phases of Pg, Sg, PmP, Pn, sPL. The result shows the two above methods produced very close error estimation. For the epicenter in the upper crust, on the condition of travel time error within 0.1 s, in order to ensure the depth error is less than 3 km, we should select these direct wave phases recorded within epicentral distance of 30 km to locate. If the travel time error is up to 0.2 s, we should select these direct wave phases within 20 km to locate on the above same condition. When the hypocenter in the lower crust, the limit of epicentral distance could be broaden a bit. As the epicentral distance or travel time error becomes larger, the error of depth location becomes practically several fold increase. Whereas, the seismic phases of PmP, Pn, sPL can make a better error constrain when the hypocenter in the upper crust, and without an obvious enlarging effect on error as the epicentral distance increases. For these three mentioned phases, they also can ensure the depth error within 3.5 km when travel time error is set within 0.1 s and epicentral distance is less than 90 km. Furthermore, by the chessboard mode we analyzed the quantitative effect of travel time resulted from the velocity disturbance. And based on the capital seismic network, we analyzed the improvement of depth location after adding head wave phases. The above analyses result show that, within 2% velocity disturbance and without simultaneously too large or too small velocity deviation for the lower crust and Moho interface, adding head wave phases can effectively produce a reliable focal depth within 3 km, and also produce a more homogeneity result than only direct wave phases used.