Hi.  I have had this idea for a roll reader, that might be simple to
build.

There should be a way to construct an "inline" switch that could be
inserted in the trackerbar line, that would use the suction already
in the tube to throw a switch.  I would think this may require a glass
automotive fuse body with a small metal ball (similar to a ball bleed
in a Ampico B unit valve).

The small ball could have a small piece of braided wire soldered to it,
forming half of the switch.  This, in effect, would be a small valve.
A short piece of steel automobile brake line is slid in the top of the
glass fuse body; this is the other half of the switch.  This could also
be slid in or out to set the travel of the ball contacts.

The switch is installed in such a way that the balls sit on their lower
seats, and the top is connected to a manifold, filled with suction.  A
bleed may be required, which would be easy to install by 'teeing' the
bleed back into the tracker bar side.

As the rush of air is detected, the balls should rise, just like a
valve; very little travel would be required.  For a really fast switch,
the double-bleed system used on the Ampico model B might be used.

Perhaps this reader would not be as accurate as fiber optics or some
of the other great ideas here, but would be within the reach of the
hobbyist.

For this system, I would not use a 'stepper' motor.  I would wind
through the roll by hand, using the "step record" system I arrange
rolls with, one chord at a time, defaulting the notes in their re-
spective place in Cakewalk (recreating the master file).  This is the
only way I know of (without expensive equipment) of creating an
accurate master with the system I am using.

I have a MIDI file (from Wayne Stahnke, I think) that is a roll
transcription, but the notes are exactly on their measures, and in
perfect time!  That is a heck of an accomplishment!  Maybe he will
tell us how that works.

Cheers
Andy Taylor