After reading Wayne's message describing his experience with pneumatic
roll readers, I decided to describe some of my own experiments.  I have
a pneumatic reader that uses silver wire contacts on vacuum operated
switches.  It has the constant maintenance problems that Wayne
described.

 [ This is the pneumatic reader which has been in continuous service
 [ at Custom Music Rolls since the early 1980s.  If one considers
 [ the quantity of music rolls it has read, its maintenance is
 [ probably no more than that needed to keep a player piano in
 [ good order!  -- Robbie

Thinking that there might be a better way, I have been experimenting
with reading rolls by detecting the flow of air rather than detecting
pressure changes.  In this approach, the air from the tracker bar is
directed on a heated wire.

The wire is in a circuit that maintains the wire at a constant tempera-
ture by maintaining its resistance constant.  The output of the circuit
is a voltage that varies according to the air flow.  This voltage is
compared to a reference voltage to determine if the tracker bar hole is
open or closed.  The ratio of full-flow voltage to blocked-flow voltage
is about two to one.

The wire is the tungsten filament of a tiny lamp (Chicago Miniature
#6833).  The lamp has its envelope removed to expose the filament to
the air flow.  The lamp filament has a very tiny mass, and has a high
ratio of surface area to volume, so it reacts to air flow changes very
quickly.  Typical response times are less than a millisecond.

For single channel testing, I used a motorized air valve consisting of
a motor with variable speed control turning a brass disk with a shallow
90 degree notch on its rim.  The notch allows air into a tube to the
filament,  and the solid part of the disk blocks it.

The notch is also detected optically to provide a synchronizing signal
to the oscilloscope used to monitor the test.  The lamp and its circuit
could follow the on-off signals up to 30 cycles per second which was
the motor's highest speed.

I typically operate the filament in the range from 150 to 200 degrees
Fahrenheit.  The response time and ratio of on-to-off voltage are both
improved at higher wire temperatures.  Faster air flow also helps
response, but turbulence in the air flow creates substantial noise in
the output.  I typically use about one inch of vacuum to supply the
reader.  This avoids the turbulent flow problem.

This approach has no moving parts, no contacts, no bleeds, and no
springs.  It is somewhat sensitive to lint in the air flow and to
changes in room temperature.  It also takes quite a few mechanical and
electronic parts per channel.  I worked my way up to an eight channel
prototype but then I more or less ran out of enthusiasm for the
project.  Recently, I have started looking at it again.

Richard Tonnesen
Richardson, Texas