[ARSCLIST] About Project Gramophone (was Re: Article on actual recording speeds of 78rpm discs -- comments?)

[Jon Noring <jon@xxxxxxxxxxx>]

Peter Copeland wrote to the ARSC group (x-posted to PrGram):

> (2) What is the precise function of the Project Gramophone operation? If
> there is a preservation element, it may be necessary to select a standard in
> terms of bits per revolution, so different copies may be accurately
> synchronised. It would then be possible to use a digital equivalent of the
> "Packburn" first stage, choosing the quieter of the two copies from one
> microsecond to the next, and also working in different frequency bands.
> There would be little point in using 96kHz/24-bit sampling otherwise.
>     Even for web listening (with lossy digital compression), the more the
> background noise is reduced, the better it will sound at the listener's
> ears.

Project Gramophone, in a nutshell, is intended to be a repository of
high quality raw digital transfers from primarily older "78 rpm" disc
recordings. At this moment, we are thinking of primarily focusing on
pre-WWII recordings (up to the 1942 recording ban in the U.S.)

Brewster Kahle, at the Internet Archive, has agreed in principle to
archive and mirror the transfers, no matter how many we get (and we
are ultimately talking about millions of them.)

(PrGram is also looking at doing related activities, but acquiring and
archiving the raw transfers is #1 on the list.)

Until laser playback systems improve (they currently have problems
reading some discs), or high-resolution 3-D micro-topographic mapping
becomes a practical reality (LBNL is studying this technology -- lots
of hurdles), we will primarily use stylus playback (stereo cartridge).
(Although we might get one of those Japanese laser playback turntables
and work with that, too -- we may transfer using both systems.)

Besides the usual high-quality equipment (including a number of
variable size/shaped styli), and the process for cleaning and
centering the records, we will probably produce the transfers at fixed
settings:

1) Totally flat (and calibrated) equalization.

2) Preset playback speed. For discs which we do not know with absolute
   certainty what the recording speed was, current thinking is to do
   the transfers at 78.26 (3600 divided by 46). Of course, whatever
   speed the transfer is done, that will be recorded in the extensive
   metadata collected with each transfer.

3) 96k/24bit resolution.

4) Two-channel (stereo) -- the two channels will not be mixed. Having
   the information from both channels oftentimes aids in restoration.


The reason for 96k/24bit resolution, which at first appears to be
extreme overkill (from Nyquist considerations alone appears to be
a minimum three times more than is needed), is to provide more than
enough resolution (headroom) for subsequent restoration purposes:

1) Some noise reduction algorithms apparently benefit from having a
   more accurate spectral mapping of the high-frequency noise above
   the actual signal,

2) Allows for more accurate resampling in order to adjust the playback
   speed (the pitch) to the correct value.

3) May benefit in re-equalization (not sure on this one.)

4) Possibly other reasons we can't predict now. It's better to be safe
   now rather than sorry at a later time -- do it right the first time.

   As Miss Piggy said: "Too much of a good thing is just wonderful!"
   In a practical sense (other than file size), the ADC step is the
   same whether done at 48/16 or 96/24 -- may as well do it at 96/24
   and sort it out later (storage space is NOT an issue for PrGram.
   Let me repeat, do not worry about disk space!)

Note, too, that digital scans or photographs will be taken of the
label and run-out area for every side transferred, and the digital
images stored along with the extensive metadata for each transfer.
(The metadata will include every gory detail about the transfer: the
equipment used, the settings/calibration, tone-arm geometry, etc.,
etc. One cannot collect too much data. We probably will even record
the temperature and humidity in the facility logs, which can be
correlated to the date/time for each transfer. We aim to preserve
as much metadata on both the source disc as well as the transfer
which might conceivably be of use to those using the raw transfers.)

It is hoped PrGram will be able to do the transfers for free (working
on the funding aspects now). We envision two approaches: a central
facility (the first probably in San Francisco) where records are
brought in by collectors, and later a traveling van manned by
well-trained volunteers where we can visit collections (both private
and institutional) and do the transfers there. We consider it a fair
trade: "We will produce a state-of-the-art digital raw transfer of
your record, of which you will receive a copy, in return for letting
us place the same copy into our online digital archive."

Will PrGram archive raw transfers done by others? The answer is a
qualified "Yes". We will accept and archive any and all raw transfers
(and even restored recordings), but will only put the "Project
Gramophone Seal of Approval" (or whatever we will call it) on those
raw transfers we know were done within our specifications and where
adequate metadata (including label photos) were collected. I hope that
we will eventually partner with other like-minded individuals and
archival institutions worldwide, using the same rigorous transfer
process, to speed up the rate of acquiring raw transfers, as well as
to improve the specifications and technology for doing the raw
transfers.

Now, regarding public access to the recordings, that is now being
worked out by Brewster Kahle (who is currently talking with IP
attornies in Europe.) We definitely want to make the raw transfers
publicly available if we can -- maybe a library model is possible from
a "50 year" country, after making arrangements, if need be, with song
rights owners (song copyrights are usually covered by life plus 50 or
75 years of the song writer or composer, while in "50 year" countries
the recordings themselves are copyrighted for only 50 years from
either the time they were fixed or when they were first released.)

In addition, we are thinking of either a broadcast model or some other
model to make available online listenable versions of the raw
transfers. We do not plan to do high-quality restoration ourselves (at
least in the first few years -- but we are interested in starting open
source efforts to build free state-of-the-art restoration software),
but may run them through rudimentary noise reduction and conversion to
MP3 (or whatever streaming format makes sense at the time). Of course,
if someone laboriously restores any of the raw transfers and donates a
copy of the restored version back to PrGram (we hope they will!), we'd
use that for creating the streaming version (as well as possibly
making the original restored WAV file available, if we can.)

I could go on with other things PrGram would like to do, but I think
this covers some of the highlights.


Btw, we do want to assemble a Technical Advisory Board (TAB) to advise
PrGram regarding specifications and requirements -- to determine the
equipment we need and the details of the transfer process. If you have
any interest in joining TAB and technically contributing to PrGram,
whether as a private individual or representing an archival
institution, let me know at jon@xxxxxxxxxxx .

Jon Noring
Project Gramophone