In this paper, spaceborne atmospheric trace gas differential absorption spectrometer is introduced. This instrument is a new optical remote sensing instrument whose spectral resolution is better than 0.5 nm. With high resolution (spectral resolution 0.3～0.5 nm), wide wavelength range (240～720 nm), large field (114°field of view corresponds to the ground 2 600 km) features, the load is pushed and swept to achieve 1 day global coverage monitoring. This instrument acquires high accuracy UV/Vis radiation scattered or reflected by air or earth surface, and can monitor distribution and variation of trace gases (NO2, SO2, O3 and so on) based on differential optical absorption spectrum algorithm. Calibration is the premise when formally putting this instrument into operation. At the same time, in order to obtain the spectral characteristics of the load, on-ground spectral calibration is needed. According to the large field, wide wavelength range, high spatial resolution and high spectral resolution of this load, a set of spectral calibration system based on two dimensional turntables is set up. This system can finish the spectral calibration of full field of view. Spectral calibration was performed using standard spectral line method with mercury lamp as calibration source. The spectral response function is an important parameter to describe the spectral response characteristics of the spectrometer. The spectral resolution of the load can be obtained according to the spectral response function. It is also the key input parameter of inversion which is based on DOAS method. The accuracy of the spectral response function directly affects the inversion results of the atmospheric trace gas. According to the spectral response data of load tests, three function models of Gauss, Lorentz and Voigt are selected as the potential spectral response functions. In order to find the most suitable function model, two kinds of contrast tests are carried out. First, the Gauss function, Lorentz function and Voigt function are used to fit the monochromatic light response data of the load, and the sum of the squares of the three kinds of functions is used as the evaluation criterion, the fitting results show that the sum of the residual squares of the Gauss function as the slit function is 0.01, and the sum of the residual squares of the Lorentz and Voigt functions as the slit function is 0.033 and 0.021 respectively. From the analysis of the fitting results of monochromatic light response data, the Gauss function could be used as a spectral response function model of load. In order to further verify this conclusion, DOAS inversion of NO2 experiment was carried out, and the influence of three kinds of function models on inversion was investigated. The NO2 sample gas test was carried out in the laboratory. The atmospheric scattering light was incident through the 30 cm×40 cm quartz window to the load slit, and the NO2 sample pool was placed in the middle of the load slit and the quartz window. The data obtained were NO2 like gas absorption spectra, and then it was filled into the N2 gas to obtain the reference spectrum of the inversion. The experiment was carried out in sunny weather and can be completed in a short time, which can reduce the influence of weather conditions on the inversion results. In the experiment, the concentration of NO2 sample gas is 8.481 2×1016 molec·cm-2. During the inversion, Gauss function, Lorentz function, Voigt function were set as slit function respectively. The results of NO2 concentration corresponding to the different functional models of three groups are analyzed, and the function model is evaluated according to the relative deviation of the inversion results. The experimental results show that the relative deviation of the Gauss function as a slit function is 5.6%, and the relative deviation of the Lorentz and Voigt functions as the slit functions is 28% and 15.1%, respectively. The fitting results of spectral response data and gas sample inversion results show that the Gauss function can be used as a spectral response function model of load.