The estimation of the disturbance torque in a sensorless DC drive is carried out by extending the classical observer theory. Three estimation schemes are formulated according to the representation of the disturbance torque and processing of the observer states. In addition to the disturbance torque, all the schemes deliver an estimation of the motor speed. Steady-state accuracy and dynamics of the schemes are analyzed in nominal conditions, identifying the scheme with the best performance. The effects of variations in the motor parameters are also analyzed, finding that a proper modeling of the motor makes the steady-state estimation of the disturbance torque insensitive to any variation. As a test, the schemes are applied to a sensorless DC drive for both compensating for the disturbance torque and closing the speed loop. The responses obtained with the best-performance scheme are reported. >