Disabling Integral windup

In regards to choosing to disable the integral term when the control device is saturated.
Normally ,The integral term keeps adding the error amount at the set intervals (repeats per period)

When the valve or control device is at its full position (saturated)
The integral is still summing the error signal.
However if the error signal reduces a little bit ,lets say about a degree or two,
Then why wouldn’t the integral term back off a little also and allow the control device
To adjust to the new error ?
On a Johnson controller, does the integral term increase/decrease the output signal,
Or does it reset the proportional band?

Thank you

cleancondenser,

with all the respect, this is why controllers have D action.

Not all HVAC systems require D action, as HVAC systems tend to have low response and a reasonable settling time.

As my personnal experience dictates, I prefer to go with formal/built in options than build one of my one replicating an already existing function.
The time taken to build a new one and troybleshooting it (most of the times it is application oriented), I could investigate the built in option, master it and have it available in a next application, where my "custom" one may not fit perfectly. Anyway this is a personnal thesis.

Back on trail, when you say output saturated, there is always a reason for this behaviour.Unless you are quite sure that you have forseen all possible scenarios and implemented on logic (and tested) you are not sure that I disabling is worth doing. Also there are some physical obstacles as you noted "However if the error signal reduces a little bit ,lets say about a degree or two,", in controls a litle bit and lets say aproaches may drive you to unstable systems (unless tested and proven like this).

I would suggest following Ziegler-Nichols method to tune up a PID. There is also a simplified method with the same results,which i could pull out from an article and post it if you are interested...

Keep in mind not to burn out on controls and testing at any way you choose to go.Some times (most times I would say) solution is sitting just in front of you but you choose not to go with it.

On JC controllers I is winding up output, to give you also an answer and not just mpla mpla on post.

Let me know if you prefer to built your own logic, and through a couple of ideas, if you need them.

Over and out

kostasgreece said:

..I would suggest following Ziegler-Nichols method to tune up a PID. There is also a simplified method with the same results,which i could pull out from an article and post it if you are interested...

Keep in mind not to burn out on controls and testing at any way you choose to go.Some times (most times I would say) solution is sitting just in front of you but you choose not to go with it.

On JC controllers I is winding up output, to give you also an answer and not just mpla mpla on post.

Let me know if you prefer to built your own logic, and through a couple of ideas, if you need them....

Click to expand...

Anything you post in addition would be appreciated. I can see already you are going to be good for this site.