In this article a high performance Extended Kalman filter algorithm is used to estimate rotor speed, rotor position and load torque in a Hybrid Stepper Motor (HSM). Together with a PI speed controller, the Kalman filter forms a very robust sensorless closed loop system. During acceleration the motor is capable of running in closed loop from 0-100% of the nominal speed. The controller can handle charge steps of over 50% of the maximum load torque, and acceleration up to 13 [krad/s 2]. To achieve theses good dynamic capabilities, the Kalman filter is optimized by an off-line identification process. This gives a very good match between the model and the real system. The standstill position at start-up is estimated by a current injection method. Due to the optimized Kalman filter the experimental results show very good precision and closed loop robustness for a 10.5 [W] HSM. The algorithm is implemented on a floating point DSP. The driver consists of two H-bridges with MOSFET transistors. The currents are measured with one shunt resistor for each phase.