the field of motion control, robustness is an important issue to accomplish desired motion of robots. To achieve robust motion control, acceleration control has been studied. In the acceleration control, disturbance observer is used to estimate and compensate disturbance force applied to an actuator. In the disturbance observer, the disturbance is estimated with calculating acceleration based on velocity/position information, and the estimated disturbance is obtained through a low-pass filter. To achieve accurate disturbance estimation, higher cut-off frequency of the low-pass filter and exact acceleration information are required. To set higher cut-off frequency of the low-pass filter, shortening sampling period is an effective method using FPGA as a motion control processor. However, short sampling period generally deteriorates the resolution of the velocity/acceleration information. This means that the estimated disturbance can include large error. In this paper, an acceleration estimation method is proposed to improve the resolution of acceleration in systems whose sampling period is very short. The validity of the proposed method is demonstrated by simulation and experimental results.