Dealing with Deadtime

Process deadtime is troublesome for a feedback controller, but there are ways to work around it.
Vance VanDoren

AT A GLANCE

• Deadtime examples
• Problems caused by deadtime
• De-tuning the controller
• The Smith Predictor

For feedback control, deadtime is the delay between the instant a controller begins to apply a corrective effort to the process and the instant a sensor first notices an effect. The process variable cannot react at all to the controller's efforts during the de

adtime, and any attempt to manipulate the process variable before the deadtime has elapsed inevitably fails.

Consider, for example, a process consisting of a car with a loose steering wheel. The driver controlling the car may try to make a turn but will sense no change in the car's direction until the steering wheel has moved far enough to overcome the slop and engage the steering column. The time required to do so is the deadtime.

Deadtime is arguably one of the most difficult control problems to overcome. In the above example, a driver unaware of the deadtime in his process will begin to steer even harder in the direction he originally wanted to turn upon realizing that his first attempt has had no effect.

But by the time the deadtime has elapsed and the car begins to turn, it will be too late. The driver will have overcorrected and the car will end up turning too far. The driver will then have to start turning back in the other direction and will eventually lose control of his process.

 

Incidentally, this same phenomenon is responsible for so many drunk-driving accidents. The car's steering wheel may be perfectly responsive, but an inebriated driver cannot perceive that his car is turning until it has already been doing so for some time. In this case, the lack of control is due to deadtime in the human sensor rather than the process, but the results can be just as catastrophic.

Sensor location, patience

In either case, eliminating the deadtime is the obvious solution