[vorbis] 48000 Hz in vorbis rc3?

[Moritz Grimm]

Moz wrote:
> > The higher sampling rate isn't there to reproduce
> > high frequency signals (which are inaudible), it's
> > there to reduce aliasing.

... aliasing that becomes only audible if you choose to further maul the
sample in editors. I suppose. I'm working with samples at both 44.1Khz
and 48kHz for years now, and I could never figure out the difference.
Those few kHz (almost) don't matter at all. The 16bit quantisation noise
is far worse, I wish I could work with 24Bit or 32Bit samples. IMO, the
additional overhead makes sense when it's big enough, like in 96kHz.

> Or, in less technical language, it makes a real difference to how we
> hear sudden noises like claps and cymbals. One of the trivial
> demonstrations of this is to record a handclap on DAT then play it
> back cut off at 22kHz (DAT native mode) and, say, 15kHz. Almost
> everyone can hear the difference clearly, even those who can't hear
> continuous 15kHz.

I once unsuccessfully ABXed a music clip, which I brickwalled at 16kHz
(20th order lowpass). But then, I also didn't try hard, I just wanted to
figure out whether my mind was playing tricks on me when I listened to
the clips while knowing about the who-is-who (which was true, btw :) ).

But then, I also doubt that I'd fail to ABX anything brickwalled at
15kHz ... I should try that, though I'm very sure about it, especially
if it's a problematic sample (like a really well-sampled clap or so). I
think the 48kHz of the original wouldn't matter in this case.

More interesting should be, to take a 48kHz sample and filter it. Then
to copy the filtered version and downsample the copy to 40kHz. Both
samples are filtered, and one of them is 48kHz, the other 40kHz (as
sample rate). I am sure, even 8kHz difference don't matter if there are
no more high frequencies to represent.

If you can ABX that without pitching the samples down 3 octaves, I'll
just shut up, ok? :)

<p>Moritz

P.S.: One more thing about 44.1kHz vs 48kHz - playing back a sample of a
C-4 at 44.1kHz and pitching it up to 48kHz doesn't even make it a D-4
... it's somewhere between C#4 and D-4. A difference of 3900 Hz is
marginal at these high frequencies. On the other hand, 96kHz DOES make a
difference. Compared to a 44.1kHz sample, you can go up ~13.5 half tones
(12 half tones is one octave, which is what you can go up compared to a
48kHz sample). I suppose this is exactly the reason why Vorbis performs
so well on 48kHz samples with its modes tuned for 44.1kHz.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.