It looks like I missed 970415, Jody.  I never saw this message:
[ Look in 970416  ;)   -- Robbie ]

Hi, all.  In MMDigest 97.04.16, Beatrice Robertson wrote:
> And Larry, there's another project for you.  You would sure be a hit in
> MBSI if you made an adjustable microswitch mechanism that could be made
> to fit many music box comb layouts!  That would be a standing-room-only
> workshop for a convention!

As it happens, I've been thinking along these line.  When I suggested the
above I was thinking in terms of very small microswitches arrayed along a
rod, each secured in position by a small screw.  Nothing too complicated.
But adjusting such a critter would be a major pain in the posterior.
Since then, though, I had couple of other ideas.

The first idea is the simplest: dismount the cylinder, ink the pins tips,
and roll it across a sheet of paper.  Since the tips are, theoretically,
all the same height, this should leave an accurate record of all the
dots, ready for scanning.  This seems rather obvious, but I don't recall
hearing it suggested previously.

The second idea is to use a pad like those used for mouse input on modern
laptop computers, but arranged in a thin line instead of pad -- that would
give you the equivalent of a few thousand microswitches in a row.  Mount
it in place of the comb, spin the cylinder, and a computer ought to have
plenty of info to figure out the tooth spacing of the comb and the
arrangement of the notes. It would be a good idea to layer a piece of
disposable plastic on the front of it, since the pin tips will likely
score it no matter how lightly we arranged for it to press, but these
things are sensitive enough that I'm sure they could be detected even
through such a "sacrifice" layer.

The third idea is to use a sensor and a stepping motor on a shaft.  With
this idea, the sensor -- it could be an infrared laser and photocell, or
something as basic as a single microswitch filed down to the thickness of
a pin -- is mounted on a shaft and placed at the first pin track.  The
cylinder is spun, and the sensor recorded.  The stepping motor then
pushes the sensor to the next track, and the process repeats.  In it's
simplest mechanical form, the "stepping motor" is just a person adjusting
a knob to the next track.  In its most sophisticated form, a stepping
motor or voice coil would move the sensor to the next micro-track -- which
might or might not be the next physical track.  When the cylinder spins,
the sensor will read out a strip that either matches (in whole or in
part) the previous strip, or reads nothing (between actual physical
tracks) or a new strip (meaning it just reached the next physical track).
Either way, it would be fairly easy to disambiguate this in software.  A
non-physical scanning system -- say, an infra-red laser and a spinning
mirror -- could actually do horizontal scan lines and capture the cylinder
in one pass, resulting in a scan file rather like the end result of the
paper method outlined above.

Hope that sparks some ideas...

regards,
Larry Smith