Acceleration feedback techniques have been used in control systems for a long time in order to improve the stiffness. Basically, the acceleration of the motor is calculated using speed or position measurements and the current command is adjusted using the acceleration data for avoiding deviations from the commanded speed. This method has to deal with two challenges. The first challenge is calculating the acceleration correctly. If it is calculated by double differentiation of position feedback of a servo motor, it may lead to a high amount of noise. On the other hand, if observer techniques are used, the error and variation in system parameters and quantization noise in the measured current may lead to incorrect calculation of the motor acceleration. The second challenge is avoiding performance degradation in the transient response of the system and preventing oscillations. Since acceleration feedback will try to avoid any acceleration in the system, it may severely affect the transient-state performance. This paper presents a disturbance rejection method that does not depend on system parameters and that does not affect the transient-state response of the system. The method provides very significant improvement in disturbance rejection over a wide frequency range. The magnitude of disturbance response was --29.6 dB in the original scheme at its peak frequency of 5.17 Hz; using the proposed method, it improved to --44.7 dB at the same frequency.