All this roll reading correspondence has been most interesting.  I get
the feeling that there is real progress, and look forward to having my
own roll transcriber one day!  In the meantime, a few ideas of my own...

Jon Miller's question about the need for accuracy in roll transcrip-
tions suggested new philosophy, applying Le Corbusier's 'less is more'
to the player piano as 'worse is better'.  The bad rebuilder's charter:
many will leap at adopting it!

If you sense the actual roll perforations when making a software copy,
you have 100% of the detail on the original roll.  You can then choose
at playback to emulate any level of imperfection in the instrument --
sticking notes, poor valves, whatever!  The possibilities are endless.

Even if you wish to convert a roll to MIDI, think about it.  Say you
can transcribe at 100 lines per inch.  You convert results straight to
MIDI, so you have an error of up to a 200th of an inch either way on
every note.  But, recreate the original perforation master and you can
create a MIDI file with zero error on every note, which must be better.

Taking this further:  When using a re-created master to cut new rolls,
you have to consider how the original and new perforators work.  The
convention has been that cutting at the nearest 50th or 100th of an
inch is 'good enough'.  You end up with rolls of the same length but
with slight timing errors on each note.

Here in the UK, Steve Cox's 'Laguna Rolls' perforator steps in
increments of 200 per inch (2400 per foot), the software selecting
the necessary number of such steps per punch row.  Such a design could
emulate the step advance of original cutters, for instance, cutting
with 4-step rows from a re-created master would give 600 rows per foot,
which is not too different to Welte Licensee's 540 per foot.  (Not
close enough, perhaps, but it's just an example).

This approach gives old and new rolls of slightly different length but
identical pattern of holes, so each would produce the same performance
if the roll tempo was adjusted accordingly.  (This happened with some
original Duo-Art rolls where the UK perforator step advance was differ-
ent to the USA one, so the marked tempo was adjusted to compensate).

A difference in lengths obviously affects acceleration characteristics,
so length differences should be kept low.  The point is, it would be
better to have any slight error in the roll length rather than in
individual notes.

I still think that the best way to preserve an old roll is to make a
new roll.  Computers have a distressingly short life, and archiving
electronic data is something that even big companies and governments
cannot get right.  It would be very sad to lose all the electronic
rolls after their paper progenitors had crumbled to dust!  So, all
those making transcribers need to be matched by equivalent roll cutting
projects.  Who will take up the challenge?

Julian Dyer

 [ If a perfect recut can be made (which is easily checked against the
 [ original roll), then a perfect computer file or MIDI file is also
 [ possible.  Both markets are served with this philosophy!  -- Robbie