Assume input signal and discrete output. The signal can cross the limit at 1000 times per second, but it is desired that the discrete output changes at the most once every 15 seconds. See image below:

enter image description here

$U$ is the Signal, $A$ is the comparator output, $B$ is the filtered output. The goal is to filter out the output changes in region $R$.

  • Are there theories to dealing with volatility around the limit such as in region $R$ ?

  • At this point I am thinking of an algorithm where state is only changed if the signal is persistent for time period $T$, is there a name for such an algorithm?

  • 1
    $\begingroup$ Do you mean that the state of the system determines the which dynamics is used? So for example $\dot{x}=Ax+Bu$ and $u=\max(-1,\min(Kx,1))$. $\endgroup$
    – fibonatic
    May 23 '16 at 23:40
  • $\begingroup$ Actually the other way around, want to reduce frequent changes of state due to volatile signal. The signal can cross the warning limit at kHz, but the state should only change from normal to warning and vise versa at most one per 15 sec. $\endgroup$ May 24 '16 at 0:03

In electronics, a Schmitt trigger converts an analog signal to a digital signal by using two different thresholds (limits as you call them), one for an output state change from 0 to 1 and another for a change from 1 to 0. The first is somewhat higher so that if the input fluctuates between those two thresholds, then there will be no change in the output. The general concept of reluctancy to change state is called hysteresis.

It can also help to lowpass filter the signal first, which is often used in debouncing a mechanical switch, to remove fluctuation very much like in your picture.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.