An adaptive neural speed controller of a dc motor is proposed. The artificial neural network (ANN) is trained by the online backpropagation algorithm. The output of the ANN gives the control voltage applied to the dc motor. The difference between the reference and the actual rotor speed of the motor is backpropagated through the ANN at each step of the control process for updating the connection weights of the ANN. The control scheme requires neither a knowledge of any motor parameters, nor preferential training of the ANN. The performance of the controller is simulated depending on the rotor speed noise of the motor, the rapidity of its dynamics, the sampling period, and the sharp instantaneous change of the load, or in the reference speed trajectory.