naturerecordists
[Top] [All Lists]

Re: some respect for ATRAC (sample rate changing)

Subject: Re: some respect for ATRAC (sample rate changing)
From: "Raimund Specht" <>
Date: Tue, 29 Apr 2003 14:07:02 -0000
--- In  Rob Danielson <>
wrote:

>I've never detected artifacts by ear with header changes. I wonder
>if one compared two sound files, one created with a header change,
>if that would test the spectrogram test?

Rob,

You are fully on the right way. A header manipulation will never
change the sound quality (not even the smallest details). This is
because you do not touch the binary data at all. You are just
manipulating  the sampling rate attribute of a sound file. As you
noted, this would only lead to slight speed and pitch variations if
you play such a file at that different rate. So, there is no need
for a spectrogram test to proof that loss-less manipulation.

Regards,
Raimund

>
> aaron wrote:
>
> >  > of the artifical test signal from 44.1 to 48 kHz and then
> >>  transferred the data. The resulting spectrogram shows
incredible
> >>  artifacts:
> >
> >FWIW in the studio world, one can spend many thousands of dollars
on
> >dedicated hardware to do high-quality resampling; the software I
use
> >(Samplitude)  has a 'high quality' mode that can take an
*extremely* long
> >time to do best-quality (nearly artifact-free) conversion...
> >
> >...the received wisdom is, if you're ever going to deliver on CD,
record
> >at 44.1.  There's a running debate about whether 88.2 is a better
choice
> >for high-sample rate recording than 96, specifically because it's
easier
> >to drop every other sample than resample...
> >
> >...since I record on MD at 44.1, it's never been an issue for me.
> >
> >best,
> >  aaron
> >
> >   
> >   http://www.quietamerican.org
> >
>
> I have no clue as to why the results of sample rate changes are so
> unpredictable but, because of this, I stop to consider whether
> resampling is necessary and sometimes, it's not.  The Header
change
> command in Soundhack (mac) converts a soundfile's sample rate
> instantly with no processing time, resampling or quality change
other
> than the change in pitch and time (with 48K-->44.1K, a drop in
pitch,
> stretch of time of ~9%). That's considerable, but with broadband
> sounds (those with low pitch formation like a waterfall or
pronounced
> "air" in a location recording) the result from a header change can
be
> more pleasant than resampling. Occasionally, two header changes
can
> replace the need for resampling. My  DAT and DV camcorder record
at
> 48K, most of my library of sound files were saved at 44.1 for CD
> outputting but my DVD/DV aps need 48K.  I change the header of the
> DV/DAT sound files before importing them into my sound editing ap,
do
> the mix and then change the header of the mix back to 48K and pull
it
> into my DV ap for adding video.  (Sweetening during the sound mix
> does not create a problem because the overall tonal balance/pitch
> relations are preserved even as the final pitch/time of the mix is
> changed.)
>
> I've never detected artifacts by ear with header changes. I wonder
if
> one compared two sound files, one created with a header change, if
> that would test the spectrogram test?
>
> Rob D.
> --
> Rob Danielson
> Film Department
> University of Wisconsin-Milwaukee



________________________________________________________________________
________________________________________________________________________

<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU