Without wanting to beat my own drum, I thought it pertinent to
contribute to the discussion about preserving piano roll music by
explaining what I am doing towards this end.  In short I am undertaking
a Ph.D. at the Sydney Conservatorium of Music, my topic being the
preservation of piano roll recordings.

I am aware that many people have contributed towards this goal, using
the medium of roll scanning.  My approach is somewhat different, in
that I have developed a sophisticated unit I call a 'roll reader'.
This device 'reads' the roll in the same way as in a player piano,
with the passage of air.  The difference between this methodology and
scanning is the type of MIDI file that is produced.

In the case of roll scanning, the resulting MIDI file is of the roll
data.  This file can then be used with a MIDI-controlled perforator
to produce duplicates of the roll.  A roll reader produces what I refer
to as a 'performance' MIDI file.  The major difference is due to the
travel distance of a perforation across a tracker bar hole.

In principle, the functions in a reproducing piano (e.g., notes,
pedals, expression) are controlled by pneumatic valves.  These valves
are triggered by the roll passing over the tracker bar.  The output
of these valves is therefore the performance data that is then acted
upon by the player mechanism.  Because a roll reader incorporates the
equivalent of pneumatic valves, the electrical output from these valves
(or sensors) is therefore performance data.

The advantage here is that files thus produced can be used in a number
of ways.  The first is with a MIDI-valve equipped original instrument.
When recording rolls on my roll reader, I monitor the recording on
an original instrument fitted with the late Bob Hunt's Virtual Valve
system.  I can do this, as the data is correct for playing on such a
setup.  For example, Duo-Art has the expression holes in the tracker
bar offset from the holes for playing notes.  This ensures the
expression is set before a note strikes.

A MIDI file from a roll scan will not show this difference, whereas
a roll reader file will.  In fact, the three major systems (Duo-Art,
Ampico and Welte) all have certain aspects associated with their
operation that requires a MIDI file of the roll that is derived in the
way I mention.  Of course it is entirely possible to convert a MIDI
file from a roll scanner to a performance file, but there are many
issues to consider that could lead to a reduction in the accuracy.

Another application of such files is playing them on a modern MIDI
piano.  This requires a MIDI file of the roll performance (the eRoll
file) to be emulated.  This is a topic in itself, but suffice it to say
that a good starting point in this process is having a MIDI file of the
roll performance, not the roll data.  My thesis, when it is written
will address all these many aspects.

I strongly believe that piano roll recordings have a lot to offer.
Hence, at age 67, comes my desire to explore this through a Ph.D.  My
interest in piano rolls began in 1976, and perhaps forty years later
I will be able to contribute something towards their preservation and
increasing the interest of others in these historical performances.

I have so far recorded into computer over 5000 piano rolls, comprising
Ampico, Duo-Art and all three types of Welte rolls.  It will go on
for years more, and my hope is to make it easy for anyone wanting to
research piano rolls, or simply wanting to listen to piano roll music.

A comparison is transferring acoustic recordings from wax cylinders and
the like to CD.  Until this was done, it was difficult for anyone to
hear these historical recordings.  So too with piano rolls.  Their day
is not yet done, and I believe the true significance of piano roll
recordings has yet to be fully realised.

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