Based on the analysis of error convergence in digital learning control systems subject to measurement noise and input saturation, a new practical design method of learning control system is proposed which achieves tracking for uncertain systems through the repetition of trials over a fixed time interval.