Re: [ARSCLIST] Optical Groove Digitization

[Jon Noring <jon@xxxxxxxxxxx>]

James L. Wolf wrote:

> Just wondering. Once the processing and storage capabilities for 3-D
> modeling of a groove arrive (10+ years?), wouldn't it also be possible

Personally, I think we are there already. Disk space is not an issue,
so we can collect and process huge amounts of information, a prequisite
for doing full 3-D topography of a record to the resolution needed. And
as I noted, it would surprise me if there isn't already a surface
topography measuring system or device (probably using lasers and
suitable for our needs) being used right now in the research
establishment for some purpose. It's amazing what technologies are
routinely being used at the National Labs (and many industrial labs),
many of which have not been commercialized. Even if the specific
device hasn't yet been developed, there are probably people (the trick
is finding who) at the National Labs or at universities, who could put
something together rather quickly. I don't see this as being a big
issue, having worked in the National Labs for almost 15 years.

To me, the biggest issue is writing the code, with the requisite
computer algorithms, to analyze the 3-D topography and therefrom pull
out the recording -- to "virtually" play the recording and derive the
best possible signal from the topography of the grooves. This will NOT
be trivial. The one nice thing is once a 3-D topography is stored and
archived (probably on MO disks), then as the digital "play-back"
algorithms continue to improve over time, we can revisit earlier
topography data without need to re-scan the rare original record.

In addition, the precision of 3-D lithography continues to improve
(which is used to make intricate physical objects from digital 3-D
data), and it may be possible to "produce" a new physical record using
the 3-D topographic data if mechanical (by stylus) playback is needed.

Thus, this suggests the first stage is to develop the 3-D topographic
measuring technique, and then apply it now to extremely rare/valuable
recordings (even ones which are completely broken -- the full record
can be digitally stitched together.) Then, the next step will be to
begin developing the computer algorithms to digitally "play-back" the
3-D topographic data and restore the recordings. I foresee the digital
playback algorithms to be in a continual state of improvement and
refinement over time (and also for the 3-D topographic mapping, but
I think refinement of that will come quickly, mostly in being able
to do it faster.)


>    From what I recall of the lecture at the Library by two guys from
> Berkeley who are working on something similar to the folks on the web
> page cited, creating a 3-D map of a groove doesn't require anything
> physical to be placed in the groove (if that's what you mean by a
> probe), just two linked "cameras" to plot the coordinates of each point
> of the groove. But it takes a hell of a long time. I think with current
> hardware and technology it takes a couple days to 3-D map a 2-minute
> cylinder, maybe more.

Oh, definitely. Using any mechanical system to ride in the grooves
is totally unnecessary. We only need to make a 3-D digital "snapshot"
of the recording's surface. The key is to be able to do it to high
enough resolution with minimal error. There is the issue of speed, but
I think over time the system can be perfected to run quite fast.

As an aside, there is work now being done on 3-D memory/data storage,
which requires two or three laser beams to focus on a point within a
3-D matrix of some sort to determine if it is a '0' or a '1'. They are
talking about sugar-cube sized chucks of this material holding 50
gigabytes to a terabyte of information, and that the transfer rate can
be as high as 1 gigabyte/second. I think a laser scanning system for
precision 3-D topographic mapping of records can eventually be
designed to run at very high speed, where mapping the entire surface
of a 10" 78 rpm disc could be done in just a few minutes, if not just
a few seconds.


>    But 3-D is the only way to go. Jon Noring is absolutely right; 2-D
> is a waste of time. It only reads the edge(s) of the bottom of the
> grove, the results I've heard  (under relatively good listening
> conditions) were really poor, and vertical grooves are impossible, so
> there's no point in messing with it, except maybe for emergency
> preservation of broken laquers or something similar.

Putting it another way, when one drags a stylus (or even a laser
beam) through a groove, one ends up linearizing all it picks up. It
is a *huge* loss of information. I think that by avoiding this high
entropy-producing linearization, we will be able to filter out a lot
of the noise and distortion of older recordings *before* they are
digitized. Once you mix the noise with the signal, as what happens
with linearization (linear playback by stylus or laser), it is nigh
impossible to reverse this process (truly separate the noise from the
wanted signal.)

It would not surprise me that for some records, using the 3-D
topographical mapping approach, and developing advanced "playback
algorithms", we will see miraculous restoration (much better
signal-noise ratio and much less distortion) now impossible to
achieve by the current state-of-the-art transfer and digital
restoration methods which rely on dragging a stylus (or a laser beam)
through a groove and then trying to untangle the huge mess it
produces. This is not to say that further digital sound processing
restoration of the signal will not be necessary, but that what is fed
to the final stage of digital sound restoration (declicking/
decrackling/noise reduction/etc.) will be of much higher quality,
maybe approaching vinyl test pressing quality even from shellac
sources in say E- or E condition.

Jon Noring