More on external-reset feedback
I’ve read Greg McMillan’s articles and columns for years. When an end-user client of mine mentioned they were replacing a valve in the vapor recovery unit (VRU) and asked for my input, I laughed and mentioned the common mistake of using on/off valves as posers for control valves. [See "Tuning or valve responsible for swings?" Aug. '21, p. 46]. I only knew about this bad practice because it's a pet peeve of Greg’s.
The valve was originally intended to be an on/off valve used with on/off control logic. The OEM of the VRU did the valve installation and logic. When the level control didn't work well, they decided to change it. In a meeting, the end user and OEM mentioned the plan to change the valve. I relayed what I knew from your admonitions. But I had no idea they'd actually do it!
My tuning improved the control to the best possible under the circumstances. The level no longer gets out of whack, but the problem of the wrong valve and attendant oscillation, premature wear, and diminished life of the valve is not going to be rectified.
While I have your attention, I do have a question. You've discussed external-reset feedback several times in articles I've read over the years, but the concept still eludes me.
Does using external-reset feedback mean the same thing as preventing the integral term from accumulating more when the CV reaches zero or 100? If your answer is yes, then perhaps I couldn't find it because its implementation is built-in to my controller. If your answer is no, then I'll have to delve into it more to understand it.
You also mentioned a limit cycle. When I looked that up, all the results are math-based. So let me ask you this. In your statement "...backlash causes limit cycles...," would your meaning be the same if you had said backlash causes severe oscillation?
William Love
[email protected]
The author responds
A limit cycle is an oscillation that has a constant amplitude that never goes away while the PID is in automatic. It can be hard to spot due to other things going on like data historian compression, noise, disturbances and setpoint changes. Backlash dead band and stiction resolution can cause limit cycles if there are two or more and one or more integrators, respectively, anywhere in the system.
To my dismay, integral action is being turned on by default in modern day positioners resulting in the need to reduce the positioner gain and add integral dead band to stop oscillations from backlash. The original pneumatic positioners were high gain, proportional-only controllers.
External-reset feedback isn't commonly understood, and while it was in all pneumatic controllers, it's probably only available today in Foxboro and DeltaV controllers due to a lack of understanding of the value of the positive feedback implementation of integral action.
It's interesting that we lost some knowledge about PID controllers and positioners from the pneumatic days. Now in draft form, the ISA 5.9 Technical Report on PID Algorithms and Performance includes my Section 6 on External-Reset Feedback. I also wrote Section 7 and Appendices B through E and signal selection in Section 8 and the benefits of signal characterization in Appendix A. If you'd like to join the committee to review and comment on this report, I can add your name to the list.
Greg McMillan
[email protected]
Corrections
The news story "Inductive stages 2nd online event" (Oct '21, p. 45) contained three errors: Inductive CEO Steve Hechtman's name was misspelled, there was no first reference for Travis Cox, co-director of sales engineering at Inductive; and BHP's Firebrand award-winning application monitors 1.1 million data points. Control regrets the errors.
