High-Accuracy Automatic Calibration of Resolver Signals via Two-Step Gradient Estimators

In the software-based resolver-digital converters, angular position and speed are often demodulated from resolver signals after envelope detection. The accuracy of demodulation depends on the degree of perfectness of two envelope signals of the resolver. However, the envelope signals cannot be derived perfectly from the resolver outputs and they are often disturbed by amplitude deviations, DC offsets, and nonorthogonal phase shifts. In order to attenuate the effects of the unexpected factors on the accuracy of demodulation, an offline automatic calibration strategy is proposed for the two envelope signals of the resolver based on two-step gradient estimators. The first-step estimator is designed to obtain the estimate of only one parameter, i.e., angular frequency. At the second step, another gradient estimator is designed to estimate amplitudes, DC offsets, and phases of the two envelope signals. Since the angular frequency is estimated by an individual estimator which is independent from the other parameters, its estimation accuracy can be improved effectively. By using the estimate of the angular frequency with higher accuracy, the accuracy of the second-step estimator also can be improved. The effectiveness of the proposed method is verified by simulation and experimental results.