I have recently finished rebuilding a Celestina organette, and have
been looking at updating my MIDI-to-roll-template printing program to
produce 20-note rolls.  Having hand-cut numerous 14-note tune sheets
from these templates, I was inspired by the pictures of Ed Gaida's
modified sewing machine to try to automate (or at least mechanise) the
punching process.  So I have spent many hours with ruler and eye-glass
trying to reverse-engineer my small sample of original Celestina rolls,
and I would like to ask for some feedback on the results before I go
further.

I figure the Celestina rolls travel between 4 and 6 ft/minute (20 to 30
mm/sec), with one bar of 4/4 being somewhere between 1.25 and 1.75
inches (30 to 45 mm).  From measuring the lengths of a great number of
slots and spaces, I would put the step advance at very close to 0.5 mm,
or say 48 steps/inch.  This gives somewhere between 15 and 24 steps per
beat, depending on the selection.

The shortest perforation I have found is 2.5 mm (7/64" or 5 steps),
which I presume is the actual size of the punch.  The tracker bar slots
are 1/16" long (1.6 mm or 3 steps); the notes are consistently shortened
by 1 mm (2 steps) to avoid overlapping with the next note, or by 5 mm
(7/32" or 10 steps) if the same note is to be repeated immediately.  It
looks as if the punch only strikes about every third step (1.5 mm) when
cutting an extended slot.

This is all based on a very small sample of 8 tunes from 2 rolls, so it
could easily be wrong.  I would be very grateful if any of the roll
arrangers here would like to comment on whether these figures look
reasonable for organ rolls in general, or for the Celestina in
particular, or if anyone can offer some authoritative figures.

Many thanks,

John Wolff.
Melbourne, Australia.