Interaction accuracy and transparency of force feedback devices (FFDs) are crucial in applications like remote surgery, where high force feedback accuracy (FFA) ensures the safety of delicate procedures. However, few studies have introduced the force calibration of FFDs, especially addressing the low FFA issue in high dynamic motions. This paper proposes a calibration method and a closed-loop control (CLC) strategy for an FFD to enhance its FFA. Tailored calibration models were developed by decoupling factors causing feedback force errors. The CLC was achieved by modeling the FFD's kinematics, statics, and dynamics and integrating force and current information. Experimental results show that the integration of the models and CLC significantly improved FFA, evidenced by a reduction in mean absolute error (MAE) from 0.843 N to 0.054 N and a mean relative absolute error (MRAE) from 18.89% to 1.52% in static conditions. In dynamic motions, the MAE reduced from 3.10 N to 0.370 N, and the MRAE declined from 117.66% to 22.57%. With human-in-the-loop, the CLC reduced the MAE by about 93% and the MRAE by about 92%. The ablation study showed the effectiveness of each calibration model. Our methodology can be applied to similar motor-driven FFDs.