Wayne's description of a pneumatic roll reader sounds interesting.
I considered making a similar device.  I got the idea from observing
float type flow meters.  Using a tapered flow meter tube would allow
for adjustment of the switch point to a specific flow.

I considered how many mechanical and electronic parts would be needed
to cover 100 positions, the need for a different tracker bar for each
format, the need to rely on mechanical tracking of steadily deteriora-
ting rolls and a few other considerations, and decided that it was time
to adopt a new technology.

The main consideration is the quality of the rolls that are generally
available today.  Most rolls are so warped that they can not be read
accurately unless they are run very slowly and great effort is used to
track them accurately.  The edges are usually useless for this task.

I have many rolls that are good enough, but most have some problem or
other.  They are fine to listen to while ignoring the defects, but it
is not worth the time and effort to record them, knowing that the
recording will be flawed.  With my optical scanner I have scanned rolls
that could never play on the piano, and they play fine now as roll
images.

An optical system can overcome most of these faults, and with a proper
design it would probably take less time to build it than the pneumatic
approach.  The design that I am now using, which is based on Richard
Stibbons' interface and software, is extremely simple and can be built
with available, relatively low-tech components.  We have worked out
virtually all the bugs.

Tracking mechanically and in the computer are completely reliable and
the machine can be left unattended, while scanning, without risk to the
rolls.  The raw scan data is more than adequate to reconstruct the
master roll, and excellent music can be produced directly from the
scans without reconstructing the perforator files.

Of course, anyone interested in re-creating the original perforator
master _must_ use either a stepper-driven capstan for advancing the
roll, or an accurate, high precision position encoder riding against
the paper.  A traditional, driven take-up spool arrangement would not
be anywhere near good enough.

Anyone who is interested in taking on this project and who is seriously
committed to spending the time to complete it is invited to contact me
about details.  We will offer information on building the scanner and
software to run it.

Building a scanner of this sort is not a weekend project.  It has
taken me over a year to work out the details and many long days to
build the mechanisms.  Richard has spent five years working on the
software.  Of course, the hard jobs of software development and mech-
anical and electronic prototyping are done, but it is still a big job
to build a precision device of this sort.  A lathe and milling machine
are almost necessary and certainly recommended.  It must be built to a
high degree of precision by a competent machinist, and the electronics,
although simple, must be carefully wired with attention to all the
requirements of digital circuit design.

If you are capable of doing this work or are willing to have it
contracted please feel free to contact me.  If you are not, please
discuss it with someone else.

Spencer Chase

 [ Thanks for sharing your hard-won knowledge and experience, Spencer.
 [ We talk about "reconstructing the master roll" because this is an
 [ effective way of evaluating the precision of the process.  It's ob-
 [ viously not a requirement for a transcriber, but how do you know if
 [ the system performance hasn't deteriorated?  Comparing an image of
 [ the transcribed data with the actual music roll in front of you --
 [ and observing that all the little points and scallops are captured
 [ -- is a quick and convincing proof of the system performance.
 [ -- Robbie