Comparing with infrared and visible radiation, microwave radiation has the advantage to penetrate into thin cloud. Satellite microwave sounding data make great contributions to numerical weather prediction (NWP). The main usage of satellite microwave sounding data in NWP is to assimilate into the initial field. However, the observation operator used in assimilation of satellite microwave sounding data has large simulation errors under the conditions of cloud, precipitation and complicated land surface. Besides, the error of some microwave sounding data is also very large. Therefore how to choose microwave sounding data is a crucial issue in data assimilation. To guarantee the result of assimilation and use more sounding data, in the research of AMSU sounding data assimilation, many institutions and scholars have invented quality control schemes such as scatter index and rain detection to remove data that not well simulated by observation operator. Research shows that after using quality control in data assimilation, the assimilated output is improved and the accuracy of NWP is increased. But there is no detailed analysis on the theory and using condition of quality control so far, which makes the quality control used in research institutions quite different. In this paper, the source of the assimilation error and the principle of quality control are analyzed firstly. Then the quality control schemes used at the main operational NWP center are summarized. Finally, the future development of quality control is discussed briefly.