4
\$\begingroup\$

I am a bit confused when looking at ADC's for my project. I am using an electret microphone and modifying the circuit at the bottom of page 15 here (I can't input the image itself unfortunately here): http://www.en-genius.net/includes/files/avt_030110.pdf

enter image description here

I'd like to sample audio at ~12KHz, though this will change depending on applications I'm using (testing out voice recognition). I'm going for a bit depth of 16, as the input volume will vary and probably be from quite different environments and distances.

With background supplied, I wonder if an audio specific ADC is really necessary? Example is here

I don't need stereo, only mono. These seem to be more for more robust recording and/or output situations than my application, correct? All I need is a mic input, and the amplification would already be done by the circuit I mentioned.

Also, given these values, would I simply need ~12ksps to match the 12khz sampling frequency?

Would a simple ADC that has correct resolution and SPS be fine then, supposing more of the DSP and amplification would be done elsewhere? Is there something special about audio ADC's that behaves quite differently?

Thanks for your help, I know some of these are quite elementary inquiries, though I haven't found very clear answers to them. I could be looking in the completely wrong route, so any input is highly appreciated.

\$\endgroup\$

1 Answer 1

Reset to default
4
\$\begingroup\$

Math of sampling dictates two things:

  1. You need to sample twice as fast (in the ideal, practically faster) as the maximum frequency you'd like to capture
  2. You need to remove frequencies above half the sample rate with a low-pass filter, or they will corrupt your lower frequency content through aliasing.

If you need 12 kHz, you need to sample at more than 24 kHz and provide an anti-aliasing filter before the sample is done.

\$\endgroup\$
7
  • \$\begingroup\$ Yes, Nyquist criteria. Certain voice recog. algorithms only sample at frequencies up to ~12KHz, as the 6KHz waves are about as high as necessary with upper harmonics more about how we hear things, not how a computer needs to see them. That being said, if I'm sampling at 12KHz, I need an ADC at 24KHz? Due to nyquist criteria, then, I need to double the sampling rate for each component I use? Is that what you're saying? \$\endgroup\$
    – user1569980
    Commented Mar 3, 2013 at 15:29
  • 1
    \$\begingroup\$ No, to sample at 12k, you prefilter at <6k, and have no representaion of signals above 6k. \$\endgroup\$
    – Scott Seidman
    Commented Mar 3, 2013 at 16:22
  • \$\begingroup\$ Ok that makes sense, that's simple enough. So if I'm looking to sample theoretically up to 6k, my ADC should be at least 12k? \$\endgroup\$
    – user1569980
    Commented Mar 3, 2013 at 17:10
  • 1
    \$\begingroup\$ No difference, for the base ADC, but the device you point to has an embedded DSP with some built in tricks for processing audio, as well as automatic gain control. If you need this stuff, it looks great \$\endgroup\$
    – Scott Seidman
    Commented Mar 3, 2013 at 20:34
  • 1
    \$\begingroup\$ To "So if I'm looking to sample theoretically up to 6k, my ADC should be at least 12k?" the answer is yes. \$\endgroup\$
    – Scott Seidman
    Commented Mar 3, 2013 at 20:35

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.