Compared with traditional position sensors, magnetic encoders do not require optical components, which are easily damaged, have complex winding structures, and have higher reliability, better environmental resistance, and stronger vibration resistance. However, owing to some non-ideal factors, the output angle of the magnetic encoder may have an amplitude error, phase offset, DC bias, and harmonic deviation, which greatly affect the decoding accuracy. A magnetic encoder error compensation algorithm is proposed based on an improved second-order generalized integrator (ISOGI). First, the errors existing in the magnetic encoder are analyzed and verified and then a dual improved second-order generalized integrator harmonic decoupling network phase-lock loop (DISOGI-HDN-PLL) is proposed, which has the ability to compensate for low-order harmonics and other errors. This proposed method compensates for various errors under constant and high-speed conditions. In addition, it has good response and robustness, effectively improving the decoding accuracy of the magnetic encoder. Simulation and experimental results demonstrate the effectiveness and applicability of the proposed algorithm.