This paper presents a unified approach for the design of dead-time compensators. The approach is based on a modified structure of the Smith predictor that allows to decouple the disturbance and set-point responses in order to tune the controller for a compromise between performance and robustness and is able to cope with unstable processes. The proposed structure is simple to analyse and tune. Moreover a novel robustness analysis of the unstable case is presented to demonstrate the limitations imposed by the unstable dynamics and the dead time. Cases studies are used to illustrate the advantages of the proposed controller. (C) 2008 Elsevier Ltd. All rights reserved.