Abstract: The model error of the positioning system is one of the most important factors that limit the positioning accuracy. Most positioning systems use parameter calibration or parameter difference method to reduce the influence of model error on positioning accuracy. In this paper, for the satellite radiation source positioning application, the periodic wideband signal with low SNR is used as the reference signal to calibrate the model error of the positioning system. The calibration process establishes a model error parameter direct estimation model and a direct estimation model of the radiation source position. In the model error direct estimation model, the relationship between the received reference signal waveform and the model error parameters is established, and the maximum resolution method is used to obtain the super-resolution model error pa-rameter estimation. In the direct estimation model of the radiation source position, the relationship between the model error parameter, the unknown radiation source position and the received unknown signal waveform is established, and the posi-tion of the radiation source is directly estimated. For the problem that the reference signal has large bandwidth, long period and low signal-to-noise ratio, which leads to large computational complexity of model error parameter estimation, the ener-gy accumulation and step-by-step search method of the reference signal are used to guide the reference signal step by step, which improves the system. The parameters estimate speed and accuracy. Finally, a simulation example is given to illustrate the effectiveness of the method.