The most important parameter for process control is the easiest to identify. The time after a change in the PID output till you see a significant change in the process variable is the total loop deadtime. Deadtime is directly or indirectly a factor in all aspects of control system design and tuning for loop performance. Without deadtime, I would be out of job (for startup experiences confirming this suspicion see A Funny Thing ). A control loop would see a change and make a correction immediately. The controller gain could be set as high as wanted and the control error could be as small as desired. A zero deadtime implies no filters (hence no noise) and no sensitivity and resolution limits (hence perfect measurement and valve precision) since these add deadtime to the response.
I asked Michel Ruel president of Top Control who I interviewed for Nov and Dec 2009 Control Talk columns “Show me the Money” Part 1 and Part 2 how does he set the PID execution time, filter time, and integral time to prevent excessive oscillation? Michel’s answer reaffirmed the insights learned from the great minds in Process Control (Greg Shinskey and Vernon Trevathan).
scan time < filter time < rate time < deadtime < integral time < damped period
Note that the central term is deadtime. The ultimate period, minimum Lambda (closed loop time constant), minimum integral time, and minimum peak error are proportional to the deadtime, and the minimum integrated error is proportional to the deadtime squared. For more details on the importance of deadtime for these parameters and metrics, check out the Exceptional Opportunities in Process Control Course Presentations.
Deadtime is the key for getting the feedforward timing right. In Deminar #11 on Jan 5, we see how a feedforward delay set smaller than the disturbance deadtime causes inverse response and a delay set larger than the disturbance deadtime causes a large second upset that is particularly disruptive. The tightness of the feedforward delay setting required to prevent the feedforward from doing more harm than good was surprising. If the deadtime in the path of the feedforward correction is significantly larger than the deadtime in the path of the disturbance to the same point in the process, feedforward control is not advisable unless you are using wireless transmitters and the PIDplus (Demianr #11).
Deadtime is the key to preventing resonance. Since the ultimate period (natural frequency) is largely set by the deadtime, interacting loops with deadtimes that differ by less than a factor of five, will interact if tuned for maximum disturbance rejection.
Deadtime is also the key to the short cut tuning method, the subject of a future entry on this website.