[LAU] Sample rate vs. SNR

Fons Adriaensen fons at linuxaudio.org
Wed Jan 23 16:17:45 UTC 2013

On Wed, Jan 23, 2013 at 06:52:12AM -0800, Len Ovens wrote:

> Just looking through TI's ADCs 96k and up. Any on chip audio circuitry
> (level controls etc.) seems to lower SNR (101 db typ.) which is no
> surprise. But what I did find interesting is that even 96k and 192k
> devices seem to optimized for 48k. So the spec of 112db SNR would be true
> for 48k, but by the time it gets to 96k that is down 3db (noise floor up
> 3db) and at 192k... not even listed.
> So 192k is snake oil anyway, I'm not concerned, though I will say the 192k
> device has better performance at 48K, much smoother frequency response.

Noise in 24-bit ADC is all thermal noise in the analog parts,
not quantisation noise. If it is measured unweighted it increases
by 3 dB if the bandwidth is doubled. But for 96 kHz, half of it
is above 24 kHz, so the noise within the audio bandwidth remains
just the same. There is no degradation.

The specs should tell you how exactly the SNR is measured. If the
A-weighted or 20 kHz BW noise increase at 96 at 192 kHz there is
something fishy.



A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)

More information about the Linux-audio-user mailing list