Measurement signals that a computer uses to take control actions are usually contaminated by noise. The presence of noise is undesirable because it may be detrimental to the operational control. In order to reduce the noise level in a batch distillation column control loop, two digital filters were experimentally tested: the double exponential filter and the moving average filter. The runs have shown how important the choice of the digital filter is to achieve a good control performance. For control tasks, this choice must be a compromise between data smoothing and the ability to respond rapidly to real changes in the process. (C) 1999 Elsevier Science Ltd. All rights reserved.383217224