[I'm cc'ing this to ARSCList and Project Gramophone. Art's 78-L reply
is a response to the question of why Project Gramophone should do raw
transfers at 96K/24-bit rather than the "ordinary" 44.1K/16-bit.]

Art Shifrin wrote to 78-L:

> With the exception of electronic music & other natural sounds, one of the
> significant differences between noise and content is the shape of the
> waveform of a given sample of sound.& noise.  24 / 96 provides much more
> accurate reproduction of those noises.  The more precise the distinctions
> that can be made between noise and program, then the better job of
> "lossless" ( 'minimally lossfull' is a better, non-jingoistic term) noise
> reduction could be done.  This principle is analogous to working from analog
> tape transfers of noise originals: the NOISE is more accurately produced at
> 30 than at 15, 15 than at 7.5, etc.Once I upgraded to 24 / 96, I immediately
> heard (and saw with oscilloscopic tests) improvement  in the NR functions of
> Sound Forge.  I now deliver all jobs to clients as raw 24 / 96 files wave ,
> AIFF, etc) in addition to the processed  & watered down versions that they
> require, such as Redbook, et. al. Given that many original disks and
> cylinders will continue to deteriorate (even when stored under theoretically
> optimal conditions) doing this theoretically would enable even BETTER noise
> reduction in the future, when even better software that we have now would
> hopefully be available

Thanks for your feedback. I "knew" that it is better to do raw
transfers at 96K/24-bit, even if by Nyquist frequency considerations
it appears to be major overkill for older recordings. My thinking is
that doing transfers at that sampling rate and depth gives plenty of
"overhead" for the digital restoration step which employs algorithms
that will mathematically work better at that very high resolution.

(Of course, at the end of the digital restoration, resampling to
44.1K/16-bit, Red Book specs, can be done as needed, with no real loss
in sound quality.)

> A one hour 24 / 96 monaural wave file is about 1 gig.  Therefore at least 4
> hours of extremely pure digital information can be stashed on a DVD-OR.

My current thinking is that in the field, the raw transfers done at
96K/24-bit would be stored on pc harddrives (as WAV or similar lossless
format). Up to a terabyte (on 3-300 gig IDE hard-drives) can easily be
stored this way. At a later time, the sound files can be backed up to
DVD and/or other archival media such as MO (redundantly of course.)

Now, to answer another obvious question, let me state that the
potential partner for Project Gramophone has assured me that SPACE IS
NO PROBLEM. If we generate 1 *petabyte* (which is 1000 terabytes or
1,000,000 gigabytes) of sound files (which we'd have to work hard for
*years* to accumulate), they will store them on their server. (They
are presently designing a "portable" 1 petabyte storage system which
is almost dirt cheap in capital and operating costs to run. These
people are at the cutting edge of Internet server storage technology.
And year-by-year storage capabilities continue to outstrip our ability
to fill it.)

So space is *no problem*. That is, folks, don't worry about space or
file sizes. Anyone who brings this up again as an issue will be shot
at dawn. <laugh/>

(Now, transporting a 100 meg raw transfer WAV file over the Internet
for someone to digitally restore -- that's more of an issue. However,
with broadband that can be done in minutes to an hour, which is not an
issue. The files can also be burned to cheap CD-ROMs (as datafiles and
not as audio tracks) and sent by snailmail. Thus, I'm comfortable with
this as well, so let's put all this to rest and focus on the remaining

> Given the dramatically decreased costs of AD / DA converters & storage, I
> consider it to be short sighted to digitize & store preservation masters,
> raw transfers, etc. at any less resolution.


> A 60 or even 90 minute radio program taken from disks & digitized to 24 / 96
> is now simply and easily stored on one disk, EVEN if the two outputs of a
> stereo cartridge are kept discrete for the monaural grooves.  The same
> advantage obviously applies to long duration, short form recordings such as
> complete operas, concerti & symphonies.

Again, with 300 gig IDE drives going for less than $200, storage of a
hundred hours of two-channel music at 96K/24-bit resolution is no
longer an issue.

Jon Noring
Project Gramophone