For the past year I have been working on the development of an optical
scanner for transcribing piano rolls. I began communicating with
Richard Stibbons a while ago, and from that time on progress has been
steady and at times remarkable.
Richard has developed software for receiving the output from CIS arrays
and converting the image data into music files of varying sorts. His
software can decode the peculiarities even of the Duo-Art system with
its demands of accurate timing and non-standard punch positions. He
has an Ampico himself and has covered that system as well as Recordo,
and 88 note.
All this is fed by the output of inexpensive surplus consumer sized
arrays and hardware that is embarrassingly simple. Richard has used
some very simple methods to solve some very complex problems. His
software is far from simple. It covers virtually all the nonsense the
rolls can fling at it and outputs MIDI files that are nearly perfect.
Using servo technology the notes are fitted to the positions they
should occupy if the rolls were punched correctly.
Although we have no direct interest in roll cutting at this time, we
have set a high resolution standard which ensures that our scan images
are of sufficient quality to enable the original master rolls to be
regenerated from them at any time in the future. In the last week,
Richard has been concentrating on perfecting the decoding of Duo-Art
rolls, as that is my special interest and the reason for my scanner
construction.
Being mechanically inclined, I have been concentrating my efforts on
the machine aspect. My machine is very sturdily built and should be
long lasting and reliable. It is very gentle to the rolls and will
cause less damage than playing them on a well adjusted piano.
I have really put effort into this aspect of the design. It can
scan rolls that are too frail to be played and will recover data from
rolls that are too warped to play properly, although perfect rolls are
greatly preferred. I finally have a copy of Chopin's Nocturne in
D-Flat that doesn't play false notes.
When errors are found due to serious roll damage, they can be edited
with any MIDI editor such as Cakewalk. The scanner tracks very
precisely (wander of about .010" rms) and in a fail-safe manner and
does not need to be attended. A roll is loaded and it stops when
finished, leaving me free to attend to other matters.
There are a few adjustments to be made, and the inevitable few bugs to
be removed, and we will be able to produce nearly perfect digital
images of Duo-Art rolls for all of the PowerRoll Duo-Art owners who
have been suffering from a lack of software for the Duo-Art.
I have a fairly large collection of good Duo-Art rolls and have access
to many more. I am still interested in borrowing any gems that I can't
get my hands on. Richard and I are interested in providing electronic
images free of charge to anyone who assists in the transcribing effort.
Anyone who has outstanding Duo-Art rolls in the US is invited to
contact me if they are interested in having them scanned and converted
to electronic formats. Richard is available in Europe and his interest
is primarily in Ampico. We (Richard is really responsible for this as
well as all the software) can also produce MIDI files that emulate the
Ampico and Duo-Art reproducing systems to a reasonable degree of
accuracy.
Please be aware that we are dedicated amateurs and do not have the
time to satisfy demands that are not in our interests. Please do not
contact Richard with unreasonable questions or demands as he is very
busy and is generally too polite to say no.
Spencer Chase
|