Re: arsclist cylinder frequency range

[David Seubert <seubert@xxxxxxxxxxxxxxxx>]

George et al., I'm referring to lossy compression (or data reduction if you 
prefer to call it that). My initial impression was that the Quicktime 
compression algorithms aren't able to deal with noise well. As it turns 
out, it is the Qdesign codecs that don't deal with noise well. We've 
already decided not to to de-noise, de-crackle the recordings before 
compressing them, so our codec needs to be able to deal with noise. For 
one, we don't have the money for Cedar or Sonic Solutions, and most of the 
cheaper ones I've heard sound pretty crummy. I'm glad to hear about Diamond 
cut. I'd like to give that a try.

So far we've made a couple hundred test compressions and we've learned a 
lot about what works and what doesn't. I'll load some of the more 
illustrative ones on our server after the holiday weekend. I'd be very 
interested in what other ears think of the various codecs, sampling rates, 
bit rates, bandpass filtering and all the other permutations we came up with.

David


At 09:01 PM 1/17/2002 +0100, you wrote:
> Date sent:              Thu, 17 Jan 2002 10:36:55 -0800
> To:                     ARSCLIST@xxxxxxxxxxxxxxxx
> From:                   David Seubert <seubert@xxxxxxxxxxxxxxxx>
> Subject:                Re: arsclist cylinder frequency range
> Send reply to:          ARSCLIST@xxxxxxxxxxxxxxxx
>
> David,
>
> you wrote....
>
>
>  when compressing files using Quicktime that
> > the less noise in the source, the better the compression works.
>
> ----- in which way better? - we need a terminology here.
> Compression is usually loss-less. Data reduction is not.


> Physiological data reduction (my, hopefully more precise, term for
> "psychoacoustic data reduction") actively removes that which
> might be unheard anyway by average humans because of various
> types of masking. Subsequent spectral correction may then call for
> signal components which are no longer there. "better working
> compression" might mean "enables really low data rates" or "does
> not sound too artificial", if we are really talking data reduction. At
> any rate, trying to apply signal processing to a retrieved data-
> reduced signal, in particular where the S/N ratio is low, really
> creates trouble.
>
>  Given
> > the amount of noise in a cylinder, we are going to have to use some
> > sort of low pass filter or the compressed files will sound terrible.
>
> ----- I do not think that low-pass filtering is a good thing. In the
> LINEAR world I would propose using a wide frequency range for
> distribution, because low pass filtering is so simple that an end-
> user can apply it to his or her own taste. Distributing a low-pass
> filtered signal is distributing a mutilated signal. In a data reduced
> world I appreciate that there would be a need for a different
> compromise.
>
> > We haven't yet decided what frequency the low pass filter should be
> > at. When our tests are done, results will be posted on our website.
>
> ----- here I would support Joe Salerno's idea that you might put
> something semi-manufactured on the net. If the ARSClist is not
> suitable for hefty files, then perhaps you could put it on a server,
> password "ARSClist".
>
> Kind regards,
>
>
> George
> Preservation Tactics
>
> -
> For subscription instructions, see the ARSC home page
> http://www.arsc-audio.org/arsclist.html
> Copyright of individual posting is owned by the author of the posting and
> permission to re-transmit or publish a post must be secured
> from the author of the post.

David Seubert, Curator
Performing Arts Collection
Davidson Library Special Collections
University of California
Santa Barbara, CA  93106
(805) 893-5444 Fax (805) 893-5749
mailto:seubert@xxxxxxxxxxxxxxxx
http://www.library.ucsb.edu/speccoll/pa/

-
For subscription instructions, see the ARSC home page
http://www.arsc-audio.org/arsclist.html
Copyright of individual posting is owned by the author of the posting and
permission to re-transmit or publish a post must be secured
from the author of the post.