0
\$\begingroup\$

The human ear can hear sounds up to ~ 20KHz. Some ADC's oversample. For example, to capture frequencies up to 20KHz, an ADC converter might sample at 160KHz. The steps are: 1. Filter the audio with an analog low-pass filter. 2. Sample at 160KHz. 3. Digitally filter the audio with a low-pass filter. 4. Downsample to a lower sample rate.

I understand that it's difficult to implement low-pass analog filters with sharp cutoffs: required to maximize the available bandwidth without exceeding the Nyquist limit, which would otherwise result in aliasing from folded frequencies. Therefore, I might choose to oversample at 160KHz, with an low-pass analog filter at 20KHz that can rolloff in 30KHz and still be below the Nyquist limit at 80Khz.

However, once I've sampled at 160KHz, why would I want to apply a digital low-pass filter before decimation? What are the advantages of this?

\$\endgroup\$
  • \$\begingroup\$ I suspect what you've quoted is being general. You're been specific in specifying an analog filter whose parameters may negate the need for digital filtering. & it's not that difficult to do high-order analog filters; there's chips that do it with minimal design effort needed. \$\endgroup\$ – Techydude May 4 '15 at 11:49
2
\$\begingroup\$

However, once I've sampled at 160KHz, why would I want to apply a digital low-pass filter before decimation? What are the advantages of this?

Aliasing, as a problem is not reserved purely for the analogue world - you also need to low-pass-filter it in the digital domain when decimating. The same rules apply - you need to restrict your bandwidth whenever you sample something in order to avoid aliasing or you will get exactly the same problems of folded spectrums.

\$\endgroup\$
  • 1
    \$\begingroup\$ Thanks! To summarise, we oversample because it means that low-pass analog filters don't have to rolloff in a small number of Hz: which is a win because such filters are expensive. Therfore, we can use filters that roll off over several KHz and still avoid folded spectrums. However, we then need to use a digital low-pass filter before we downsample in order to prevent aliasing when downsampling. Is that correct? \$\endgroup\$ – user54505 May 4 '15 at 13:31
  • \$\begingroup\$ That is perfectly correct. \$\endgroup\$ – Andy aka May 4 '15 at 13:52

Your Answer

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