Selecting an optimum sampling time of PID controllers

quoted from here:
The sampling operation must return the key dynamic characteristics of the process. From experience, a sampling interval of approximately 10% of the dominant time constant works well in practice.

Cicero

Hello,

there are some possibilities to delay the output value in refer to the input value. See the examples of the FAQ:

http://support.automation.siemens.com/WW/view/de/1005801

With best regards
Cicero

Marko Bursic

FB51 PID Self Tuner makes something you are asking for.
IMO if you excite sytem with step function 100% output and after reaching 50% of setpoint process value you set output to 0%. Then wait process value to begin falling. Very easy to tell, very difficult to implement.

I prefer experimental approach - you have 2 limitations for example - max cycle time = 200 ms and max overshoot = 5%.

1) Make standard OB35 implementation with amounts of PID loops you need. 4 variants: satisfying cycle time and overshoot (then optimum found), bad overshoot+good scantime (cycle to be decreased), good overshoot+bad scantime (cycle to be increased), bad overshoot+bad scantime (sorry, you need more power full CPU).

2) Increase or decrease scan time and repeat 1)

3) If your cycles have DIFFERENT constant time it's better to separate them in groups and place in different OBs. But tunning algo is similar....

Feel free to criticize

Aret

I prefer experimental approach - you have 2 limitations for example - max cycle time = 200 ms and max overshoot = 5%.

1) Make standard OB35 implementation with amounts of PID loops you need. 4 variants: satisfying cycle time and overshoot (then optimum found), bad overshoot+good scantime (cycle to be decreased), good overshoot+bad scantime (cycle to be increased), bad overshoot+bad scantime (sorry, you need more power full CPU).

2) Increase or decrease scan time and repeat 1)

3) If your cycles have DIFFERENT constant time it's better to separate them in groups and place in different OBs. But tunning also is similar....

Feel free to criticize

Aret

3) If your cycles have DIFFERENT constant time it's better to separate them in groups and place in different OBs.

Guys, i must admit i answer slighty different question. Initially you talk about optimum, i talk about any satisfying solution - ofc it's easier task, but at least we can define proper zone.

"10s in certain circumstances is better than 10ms."

True, thats why i started at 100ms - to be in the middle of range 10ms-1s:^) Didn't ever saw regulator with 10s sampling time - well, we all have drawbacks

I can't say i did very complicated PIDs, though couple times i was forced to use D-component. But still most of regulators are PI.

IBN-Service, i agree that using 10ms regulator for temperature control is useless, but didn't understood what do you mean by overfilling of I.

///////////// Warning - IMO after

Hamid, to be honest your task seems like science task, it's really interesting, but practical key is uncertain for me - no offense.... If we go back to basics and answer the question what to do with a large number of PID loops and low resources i would go for better CPU and splitting the system.

Of course, sometimes you need to find a solution without changing hardware, but think right lesson from it - how not to bump in it again rather then how to get out.