Aiming at the problem of non-ideal deviation on sampling envelope surface in current resolver software decoding, the current mainstream error correction methods are analyzed and divided into three categories: input correction, phase locked loop correction and output feedback correction. Then it is deduced theoretically that the amplitude deviation and DC deviation will superimpose a double frequency fluctuation on the output angle decoded by the software, and the orthogonal deviation will superimpose a fundamental frequency fluctuation. An input correction calibration method based on discrete Fourier transform (DFT) is proposed. The amplitude deviation, DC deviation and orthogonal deviation are extracted by this method, and the corresponding correction steps are designed for these three deviations. At the same time, it is deduced that there is a fluctuation error related to the initial phase of the envelope surface in the amplitude deviation and DC deviation extracted by DFT method when the rotational speed error occurs, and the influence of this error on the extraction results is solved by taking the average value of the fluctuation during a period of time. Finally, the correctness of the derivation is verified by simulation, and a resolver calibration experiment is set up for the towing platform. Comparing the data before and after calibration, it is shown that the output angle linearity of the phase locked loop with corrected deviation is better, the rotation speed fluctuation is smaller, and the corresponding dq axis current of the specified order is also smaller.