It is a remarkable fact that the vast majority of chemical processes are well-controlled with the PID algorithm or one of its simpler forms. A widely-accepted rule-of-thumb is that use of the derivative mode should be confi ned to applications characterized by long time delays and/or higher-order dynamic lags, and then only provided that the feedback signal is virtually free of measurement noise. According to Luyben and Luyben (1997), approximately 80% of industrial controllers are PI or P-only, with PID accounting for the remaining 20%. Other studies (e.g. Bialkowski, 1992) report a much lower PID implementation rate.As one would expect, introduction of a third degree-offreedom into the design problem enables the engineer to obtain tighter control. This is because the phase lead contributed by the derivative term reduces the impact of deadtime and other sources of phase lag which limit the achievable performance of the feedback loop. From a state space point of view, the derivative of the process variable (PV) can be regarded as an estimate of a system state. For second-order processes with no zeros or deadtime, knowledge of the PV and its derivative permits one to assign the closed-loop poles to arbitrary locations (Kailath, 1980). Perhaps less well-recognized is the fact that derivative action normally increases the robustness of a control system to high-frequency modelling errors. Unfortunately, these improvements are obtained at the cost of larger control effort, which may lead to increased wear-and-tear and frequent saturation of the fi nal control element, greater interaction with other control loops and operator distress (Harris and Tyreus, 1987).It is an established fact that fi rst-order-plus-deadtime (FOPDT) approximations can adequately explain the behaviour of a wide range of processes and many industrially-proven techniques are now available for fi tting these models to plant data (see Seborg et al., 2004). The main objective of this paper was to compareThe derivative mode is often omitted in PID control strategies because it proves diffi cult to arrive by trial-and-error at a set of constants which meet plant requirements. The primary objective of this paper was to evaluate several model-based PID tuning methods. For lag-dominant processes, it was recommended that the SIMC algorithm fi rst be employed to determine whether satisfactory performance can be obtained with PI control. If it cannot, then derivative action should be introduced using the DS-d technique. For delay-dominant systems, IMC tuning is preferred. It was observed that when confi gured with the same derivative fi lter factor, the series form of the PID controller produces smoother valve adjustments than the parallel version, at the expense of a slight decrease in best achievable performance. Increasing this parameter improves the control effort but limits achievable performance.Le mode différentiel est souvent omis dans les stratégies de contrôle PID parce qu'il s'avère diffi cile de parvenir par essais-erreurs à un ensemble de const...