I was told to be careful to use filters in the control loops as it will add extra time delays in the system and degrade the system dynamic response, but I also saw there are lots of filters in practical engineering, so how to determine the time constant/cutoff frequency of these 3 kind of filters: the setpoint filter, the feedback filter and the output command filter? Is there any guidelines or rules?

enter image description here

  • $\begingroup$ The setpoint filter has no effect on loop stability, because it's outside the loop. Using it to control overall bandwidth works well, when you have bandwidth to spare but there are unknown gains in the system. The stability is determined by the poles of everything going around the loop -- multiply it all together in s domain. Thus if you must add poles, you can choose between the forward and reverse path. This choice will impact the resulting closed loop zeros, which may (in a fair % of cases) be the determining factor in step response overshoot. So that is something that can be optimized $\endgroup$
    – Pete W
    Oct 13, 2021 at 17:58
  • $\begingroup$ some more detail on the tradeoff in the forward vs reverse path location for filter, in a previous question: engineering.stackexchange.com/a/43213/30144 $\endgroup$
    – Pete W
    Oct 13, 2021 at 18:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.