Glenn Amer's recent posting about my new roll reader has resulted
in a number of people wanting more information about it.  I'll keep
it brief, but I'll start off with why I built it.  My interest in
mechanical music is primarily in the musical legacy of the rolls.

I love the original instruments and own two of them, but I know
they will probably never play again when I am gone.  Instead, today's
mechanical pianos are MIDI based and are finding their way into musical
institutions.  So, for many years I have been interested in how best to
convert a piano roll to a MIDI file without loosing any aspect of the
performance.

A roll scanner is not suited to producing a performance file, but is
best suited to producing files to operate a perforator.  A roll reader
"reads" a roll in the same way as a pneumatic piano.  It therefore
produces a MIDI file that best suits a MIDI equipped pneumatic piano or
a MIDI mechanical piano.  The reason is simple.  A scanner, in effect,
photocopies roll perforations, and produces MIDI signals from this
information.  So a single perforation will produce a MIDI signal with
a duration equal to one perforation.

A roll reader records how the roll performs on a player piano.
A single perforation therefore plays for twice the length of time,
due to the distance it travels over the tracker bar hole.  This means
trills, fast repetition, legato and the like are preserved in the MIDI
file, but are lost in a scanned MIDI file.  On the other hand, a MIDI
file produced on a roll reader is not ideal for producing recut rolls.

My aim with the new reader was to produce what could arguably be called
archival quality MIDI files.  To achieve this has taken over five
years.  It has three main parts, the spool box, the sensors that "read"
the roll perforations, and the MIDI electronics.  Ancillaries include
a small vacuum pump to supply 3" water gage suction to the sensors, and
power supplies for the electronics.

The spool box has three motors: roll drive, tracker, reroll/brake.
The roll drive has optional deceleration, an on-board tempo indicator
in roll speak (e.g., shows 80 for 8 feet a minute), and is based on a
high speed 12V DC motor.  The tracker system moves the bar, and has one
sensing ear.  Operation can be either automatic or manual, in which
operating a switch moves the bar right or left.

The reroll unit has the motor always engaged, so there are no levers
or gears to worry about.  By applying power to the reroll motor during
play, it behaves as a brake.  A turns sensor causes the braking power
to be reduced every two turns, so the braking effect remains constant
as the roll unwinds.  The same circuit also controls the motor on
rewind, so roll speed is reduced as the roll rewinds.  My aim was to
ensure no damage could occur to rolls either during play or reroll.

The sensors were the tricky bit and took me over two years to develop,
design and finally construct.  They are similar to a primary valve in
that there is a pouch (I call it a diaphragm), a bleed and a vacuum
supply.  The diaphragm is made of neoprene rubber of 0.22 mm thickness,
available from physio shops.

The sensor units are made of 3 mm thick sections of PVC plastic 30 mm
wide strips, giving an absolutely airtight construction.  A 4 mm x
4 mm piece of flexible magnet is glued to an 8 mm high x 6 mm diameter
section of aluminium rod, with a 4 mm flat on one side.  The aluminium
rod is glued to the centre of the diaphragm, thereby during operation
lifting the magnet past its Hall effect sensor.  A bipolar Hall effect
sensor is mounted close to the magnet and arranged so a movement of
between 0.8 and 1 mm causes the device to change state.

The position of the Hall effect device is critical to achieve the
correct operation.  To do this, I built a device I call a note repeater
which has a disc with a set of holes rotating over a fixed disc with
two holes that represent the tracker bar.  I cut the holes so the
effect is to give an equal on and off time if the unit is connected to
a pneumatic player.  A small motor rotates the disc and an air pressure
of 3" WG provides the pulsing power to the underside of the diaphragm
in the sensors.

I built a special tool to hold the Hall effect sensor during soldering,
which had to be almost in the correct spot before soldering it in
place.  Once fitted, I could then fine tune the position to give the
correct response as shown by a digital oscilloscope.

Unlike my old reader built in 1978, this one is very serviceable.
Everything is modular and easily removed.  The sensor unit section can
be easily disconnected from the spool box and connected to another
spool box.  My aim too was to ensure rolls would be looked after, hence
the type of tracking I used.  I built the take-up spool from 70 mm PVC
tubing, and it has adjustable flanges and an adjustable hook so the tab
can be laid flat, regardless of the type of roll.

I am now using the reader to archive some interesting rolls, including
three recorded by Wylton Todd from Adelaide.  These were never issued,
and two are hand cut.  The Duo-Art roll was cut by Aeolian, but never
issued.  Todd's relatives are the driving force behind this project.

My aim is to record as many rolls as possible over the rest of my life.
By changing tracker bars, it can be used to record all types of rolls.
I hope to publish the design of the reader on my web site in case others
want to build one.  I am happy to help anyone in this regard.  All you
need is patience (lots of) and relevant skills.

Peter Phillips
Sydney
http://www.petersmidi.com/