I have a signal up to $3\textrm{ MHz}$. The ADC that sample it has a rate of $1.5\mu s$. So a full $T$ of the signal is $0.3\mu s$, and I can only sample each $1.5\mu s$. It sounds not enough, but I don't need to reconstruct it, but to create an envelop detector from it, over time, based on lets say $1000T$ (periods).

So, over $1000$ periods of $3\textrm{ MHz}$, I will have: $300 \mu s/1.5 \mu s=200 \textrm{ samples}$. From these $200$ samples, I need to create some envelop detector, or continues curve to then later check if it has some large amplitude changes.

  1. How can I chose where to sample the input signal so that I can get the "right points" of it- means mostly its maxes, where my sample rate is much slower then the signal ?
  2. Should I use a moving average to get this curve ?
  3. Is there another good approach expect from taking more periods?($>1000$)
  • $\begingroup$ There is no way to solve your problem. Short of using a sampler with a faster rate, you'd have to filter the signal to reduce its bandwidth to ~300kHz, and hope you can learn what you need from that. $\endgroup$
    – MBaz
    May 1, 2016 at 18:28
  • $\begingroup$ Really no way? Because 2 universities do exactly that with the same info exactly. $\endgroup$
    – Curnelious
    May 1, 2016 at 19:13
  • 2
    $\begingroup$ What's your message bandwidth, also your detector's bandwdith, and cannot you implement a BandPass sampling strategy ? $\endgroup$
    – Fat32
    May 1, 2016 at 19:52
  • $\begingroup$ @Curnelious : Do you have links to the work by those universities that does this? $\endgroup$
    – Peter K.
    May 6, 2016 at 1:33
  • $\begingroup$ @Curnelious : Is the signal a passband signal with min and max frequencies? Can you use the folding property while sampling? $\endgroup$
    – Fusho
    Jun 4, 2017 at 12:11

1 Answer 1


Is the ADC sample-and-hold fast enough to even capture any envelope peaks or even half cycles? If not, all bets are off.

If your ADC does have a fast enough capture time, then randomizing your sample times might be a good bet at capturing a few near extrema. If you know, a priori, the approximate shape of your waveform, then Monte-Carlo sampling might allow one to estimate the waveform amplitude from the statistical distribution of enough sufficiently time-randomized samples. Assuming the envelope changes slowly enough.

  • $\begingroup$ Yes varying the sample intervals can extend your reach; I am not sure random is necessary but scanning might work: say 1.5 usec, 1.48 usec..3.2 usec can peak at different signals and might clarify your signal; if it doesn't change rapidly. Something analogous to dithering: trading bandwidth for resolution. I don't know of references though. The "fast enough" can be slightly relaxed to the edges being sharp enough. As far as I know an underutilized technique is to dither the edges of pixels to peek at finer details. Of course demodulation in both cases has to synchronized with the scan. $\endgroup$
    – rrogers
    May 3, 2016 at 22:05

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.