Analog input channel settings based on transmitter response time

[williamlove]

I’m trying to improve a PID loop and am examining the analog input. I have one very specific question then a few more that would help me do this myself next time. In case it is helpful, I'm using a ControlLogix system.

1. My most specific question is, given that response time = 250millisec on my diff press transmitter, how should I adjust my RTS (real time sample rate--tells the card how often to read the trasmitter) and RPI (requested packet interval--tells the card how often to update the controller so the PID will have it)? (For this forum I will admit I'm not sure why I need both those knobs to turn and my plan is to set them the same.)

2. Is this a clearcut decision when you know the response time, or is this still a bit of an art? I am really seeking to understand the thought process or calculation process that leads to choosing the correct settings on my analog channel.

3. If you can explain how the response time relates to the choice in physical terms, that would help me in the future when it’s a different transmitter with a different response time.

4. Is response time the best parameter to use to determine my analog input channel settings? Is there any other parameter I would ask for from the transmitter provider that would better serve?

 

[djs]

There is really no reason to set either the RTS or the RPI to less than 250 msec. The data simply will not be updated any faster. The update rate on the PID instruction should be set much faster than the response of the loop. This is somewhat of an art. Faster update rate then needed on the PID instruction does not hurt anything other than increasing processing time of the CPU. In most cases this is not an issue.

 

[FrancisL]

The first question is how fast the process variable can change. Your transmitter may be able to respond in 250mS but if the process is slow (such as the temperature in a large tank) there is no point in sampling that fast.
It is the Process not your instrument that matters, provided the instrument is fast enough.
And there is nothing gained (the opposite in fact) by updating the PID faster than you sample the process measurement.

Francis

http://www.controldraw.co.uk

http://www.s88control.blogspot.com

 

[williamlove]

I am still hoping for a precise answer to my first question about the relationship between response time and RPI and/or RTS. Or to make it a less ControlLogix specific question, what how often should the analog input channel sample relative to the response time of the transmitter?

 

[IRstuff]

The point was that if your system response is substantially slower, the your transmitters might as well look like perfect step functions. Your sampling rate should therefore more reflect your process response, unless your transmitter is truly the time limiting entity.

TTFN
faq731-376

 

[williamlove]

Based on my increasing understanding, I'd like to reframe my question.

Please assume a very fast process time constant and almost no dead time.

Let’s leave loop update time and task scan rate out of this . . . I know how to use those and I know how to set them when I have specified how frequently I want to update the PV by reading the sensor.

My question is regarding how often I should update the PV based on the response time of the pressure sensor, e.g. 250millsec. I am seeking to understand how I evaluate the sensor response time and determine — in a precise way — how fast I should update the PV using the available settings of the analog input card. (Assume that I know how to use those settings when I know how often I want to update my PV.)

It would be helpful if you would explain if it is possible to update the PV too frequently, and if so, why?

 

[FrancisL]

williamlove
I said before " Your transmitter may be able to respond in 250mS but if the process is slow (such as the temperature in a large tank) there is no point in sampling that fast."
Sampling your input faster than needed just wastes processor time.
Executing your PID too fast can, for example, cause quantisation errors on the derivative term, and also waste processor time.
I really don't understand your focus on just the measurement, a control loop is a complete entity, the dynamics of the measurement and of the process act as one.


Francis

http://www.controldraw.co.uk

http://www.s88control.blogspot.com