I want to know how to determine the signal sampling rate required to reconstruct electrical signals whose frequency components are unknown, such as, for example, signals of charge/discharge voltage on a capacitor. I want to design a system capable of scanning this type of signal in real time.

  • $\begingroup$ You really can’t do that. You can only make educated guesses $\endgroup$
    – user28715
    Jul 24, 2018 at 5:00
  • $\begingroup$ educated guesses? $\endgroup$
    – AmnSyn
    Jul 28, 2018 at 19:07
  • 1
    $\begingroup$ google.com/… $\endgroup$
    – user28715
    Jul 28, 2018 at 20:19
  • $\begingroup$ I know what that means. I just don't know what criteria or specifications I should take into account to make a "educated guess" that actually works in practice. $\endgroup$
    – AmnSyn
    Jul 29, 2018 at 0:08
  • $\begingroup$ to select a sample rate, you need to know the frequency content. If you don’t know the frequency content, you have to make an educated guess. There is no general criteria for selecting a sample rate other than the frequency content of the signal. There must be a reason for sampling your signals, something you want to measure. Consider what frequency content that signal feature would manifest. You could also Google for what other people doing similar things have used. Make an educated guess. There are all sorts of capacitors. Do some modeling and simulation. $\endgroup$
    – user28715
    Jul 29, 2018 at 1:04

1 Answer 1


Sampling a signal requires that it should be bandlimited; i.e., its highest frequency should be less than half of the sampling frequency. When it's not, you should apply a preprocessing analog bandlimiting (aka anti-aliasing) filter to limit its bandwidth to half of sampling frequency by dicarding high frequency information.

Your application is about capturing capacitor charge/discharge signals which can be of spiky nature, and may posses very large bandwidths due to the spkies. If that's the case quite large sampling rates might be necessary for adequately capturing the spike events. Otherwise, if spikes are not your concern, but rather continuous charging discharging cycles are, then a much lower sampling frequency (determined by system time-constant) may be adequate.

Then very roughly, the discharging signal's bandwidth is inversely proportional to the time-constant of the discharging network, assuming a first order simplified model.

If you know this time constant, then you should select the anti-aliasing lowpass filter's time-constant to be smaller than that. This will make sure that the anti-aliasing filter's bandwidth will be larger than the signal of interest's bandwidth.You may need to refine this estimate based on what sort of signal activities are present in your sampled voltage / currents...


Your Answer

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

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