We wind up our series on measurement and valve dynamics with the timely question do we want a large time constant anywhere?
A large time constant in the measurement or valve slows down what the controller can see and manipulate, respectively. A large time constant in the process slows down the effect of disturbances at the input to the process. It gives a chance for the controller to catch up. In fact the ultimate integrated absolute error is proportional to the dead time squared divided by the process time constant. Is this the whole answer?
The process time constant must be downstream of the manipulated variable otherwise this process time constant acts to slow down the effect of the controller’s reaction to the upset similar to a slow valve. You can spot a slow valve or large intervening slow process time constant by a fast initial excursion from a disturbance followed by a slow recovery.
We have been talking about open loop time constants (time constant for an output change for a controller in manual). There is also a closed loop time constant (time constant for a set point change to a controller in automatic). We may want a fast closed loop time constant if this loop is a critical loop (e.g. reactor pressure) that doesn’t upset other loops or this loop is a secondary loop (e.g. flow) in a cascade control system. If the action of this loop upsets other loops, then you can reduce the interaction by increasing the closed loop time constant of the loop.