0
$\begingroup$

I am working with microcontroller's ADC to sample the data. I know that Nyquist criteria tells us that the sampling frequency should be at least twice the highest frequency, but it might involve aliasing effect hence producing a completely different signal. I need to capture signals at 9 KHz max frequency.

  1. Is there any general thumb rule as to what should be the optimal sampling rate (4x,5x...10x)?
  2. I also read at some forum that it's not good to oversample and at another place I read that now a days CDs, audio recorders oversample the signal. Can someone clarify?

Thanks.

Improve this question
$\endgroup$
2
  • $\begingroup$ Do you have any prrior information about the signal? (Other then its bandwidth) if the signal has 9Khz band limit then sampling it with 18khz should provide a clean restortion. There are some ways you can sample it with lower freq, for example if the signal is in passband.. $\endgroup$
    – Ran Greidi
    Commented May 30, 2021 at 5:57
  • $\begingroup$ Actually, it is an audio signal and I would be using a band pass filter to capture that part of the signal, most probably 2.5KHz-9KHz. Just wanted to know why according to you 18 Khz (twice) would give a clean restoration? $\endgroup$
    – Gagan Batra
    Commented May 31, 2021 at 6:32

1 Answer 1

Reset to default
1
$\begingroup$

Is there any general thumb rule as to what should be the optimal sampling rate (4x,5x...10x)?

This depends a lot on your application, your specific requirements and the typical spectrum of the signals. 9kHz sounds like audio, where something like 2.5x-3x would be a good starting point.

A non trivial aspect of sample rate selection is "what can the HW support easily?". In your case audios ADC are extremely good, extremely cheap and widely available as parts, reference designs (including drivers), and commercial off the shelf hardware. Unless there is a good reason to do otherwise, I would use a standard audio rate of 44.1kHz or 48 kHz. If you are worried about CPU resources you can follow this with a simple 2:1 digital down sampler and half the sample rate before processing. I would probably go with 48 kHz down-sampled 2:1 for a Nyquist frequency of 12 kHz.

I also read at some forum that it's not good to oversample and at another place I read that now a days CDs, audio recorders oversample the signal. Can someone clarify?

That's too broad a question. As with all technologies over-sampling is the right tool for some tasks but not for others. Here are some examples, where it is applied (in audio)

  1. Sigma Delta converters over-sample heavily. That's how they work in the first place
  2. Active Noise Cancelation system are often over-sampled (a lot), simply to keep the latency as low as possible and increase the bandwidth of the feedback loop.
  3. Any type of non-linear processing can increase the bandwidth and create aliasing. For example, some guitar effects processor over-sample before applying distortion.
Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks, that was a great answer. Considering standard 44.1 Khz or 48 Khz frequency is a good idea. But actually I am capturing sound from an outdoor pipe, you can say pipe sound/water sound.....would that qualify as an audio signal and hence using the above standard frequency work? If you mean the human hearing range, then the pipe sound might have frequency components above and below that range as well. $\endgroup$
    – Gagan Batra
    Commented May 31, 2021 at 6:51
  • 1
    $\begingroup$ I was going by your statement of 9kHz max frequency. It doesn't matter what type of signal it is, if you can determine the max frequency for your application, you can determine the sample rate. $\endgroup$
    – Hilmar
    Commented May 31, 2021 at 18:23

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.