Universal Computer Transcription of Any Pinned Cylinder

Dear MMDs,  Dear Robbie, John D. Rhodes, Craig Smith, Jonathan Herz,
and Leonardo Perretti:

Many thanks for your answers and postings, and please excuse me for
not having answered you sooner in this period of year-end feasts;
time goes on too fast.

When we want to "read" an orphan cylinder, transcribe its notation,
know what music it plays, and finally play the music, it is evident
from your answers that some problems are relatively easily solved,
and others are not.

Let's try and summarize :

The problems seem to be almost the same for musical box or organ
cylinders (and not really different for books of music too), as soon as
we do not know for which instrument they were made, the only sure thing
being that the instrument has not a chromatic scale.  That's why the
scan of paper rolls for chromatic pianos is easier.

As Robbie underlined, it seems the real problems are at least double :

1) transcribe on a flat surface (including a computer screen) the
pinning of a cylinder.  The principle is not too difficult (John Rhodes
gives a method), although Jonathan Herz points out the difficulties of
transcribing a musical box cylinder with some bent or missing pins (see
his posting in Dec 20th digest).  Leonardo Perretti emphasizes these
problems in his posting of Dec. 24th, trying a laser-reading of the
pins without even having material contact with them.

2) sorting out what kind of music was recorded on this cylinder.

Several problems arise then :

- How many tunes were pinned side by side ?  The study of the grooves
at the end of the arbor of the cylinder should give an answer, which
means we should not forget to scan it too !  That result should
normally give the number of keys (or teeth for a musical box).

All would be wonderful then if our instruments were chromatic -- but
that is very rarely the case !

So now a difficult part :

- Decide which keys are for notes, and which keys are for other
purposes (automatas, drums, general cancel, registers, etc.).  The
repetition and the shape of these keys should give some hints.

- Now the really difficult part : what is the scale of the instrument ?
Which tonality, so what notes are missing ?  Which repetitions
(especially for a musical box), so what notes are doubled, tripled,
etc. ?

By looking carefully at the disposition of the pins on the cylinder,
a good musician could determine roughly which part is for the
accompaniment, and which for the melody.  I had a surprising
demonstration of this a few years ago by Jan Jaap Haspels, the curator
of the Utrecht museum, in front of a barrel organ clock.

But in some instruments -- for practical reasons, as to shorten the
connections from the keyframe to some pipes, especially the big ones --
the notes are not even in their natural order, as if to confuse on
purpose the restorer !

Look at the scales in the Bowers-Reblitz book, "Treasures of Mechanical
Music", and you will see how keys for notes, and keys for other
purposes are mingled together.

So the real way with a multi-tune cylinder is probably to try and
isolate at first each "tune" after scanning, and then try to guess with
the help of a good musical computer software which notes are for the
bass, which for the treble melody, and which could be for other
purposes.

3) Then it should be possible to get something looking like a score,
which would allow to play it again on a MIDI-controlled instrument, and
try at last to assign such note to such key.

Craig Smith -- many thanks to him -- sent me an example of some results
which appear very interesting, even if the music could accept some
improvements.

He points out (see his posting in Dec. 19th MMDigest) that it is a very
difficult task -- brain consuming -- but that of course everything
becomes easier once you have recognized at least one tune !  And he
adds that it is better anyway to exercise on a computer than on the
real instrument, which could be easily ruined after so many trials.

Jonathan Herz has given a similar answer, applied to musical boxes,
especially interesting when a part of the comb is missing.

Fascinating, all this, isn't it ?!

Once all these problems are solved (!), at last we should be able
to listen to all these orphan musical box or organ/piano cylinders,
or simply preserve for future generations the existing cylinders.

My initial question was aroused by some scholar in a French
University telling me that scanning cylinders and recognizing the
music should not be very complicated.  I am able now to tell him
that it is !

Thanks to all, and please add your comments or experiences, if you
have any.

With best regards, and of course all my best season's greetings,

Philippe Rouille  (Paris, France)
http://www.musicamecanica.org/