Aiming at the problem of compensation accuracy limits, the conventional coning compensation algorithm was optimized in this work. The sampling values of the current and previous periods were used to build the coning compensation terms. The single sample, two samples and three samples were used in the current period, thus the compensation period was shortened and the accuracy limit value was improved, while the previous period sample numbers were optimized to make the compensation accuracy approach its limitation. The simulation results show that the proposed algorithm has some advantages on the coning compensation accuracy over the conventional multisample rotation vector algorithm.