[ Ref. Mechanical Music Digest Is Ten Years Old

Well, congratulations are clearly due to the folks at MMD HQ
(although I preferred its old address, Julian, Calif. -- I wonder
why?).  Speaking as an editor of a quarterly magazine, the idea of
producing a daily digest is distinctly off-putting, but you manage!

MMDigest has spawned numerous strands right across the hobby, but I'd
like to pick up one of the earliest as mentioned in yesterday's Digest:
roll scanning and copying.  This was the reason I subscribed (on 26
September 1995).  This was a topic already being discussed within
Player Piano Group circles here in the UK -- Rex Lawson had written
about it late in 1994 -- and here was a group with others interested
in this arcane subject!

Looking back, I see that in MMD 19970814 I wrote, "We have been assured
that it's possible to produce hole-for-hole copies of rolls [...] it's
clear the MMD has the collective resource to convert the theory into
general practice [...] We need roll scanning/reading systems designed
to identify individual perforations and record the results.  The file
must be portable across computer systems [...] Having a common file
format enables the use of any roll reader or perforator.  If producing
a synchronous master can be made as easy and as cheap as producing a
non-synchronous one, and the technology and ideas are freely available,
perhaps we could persuade all the present recutters to adopt the
technique."

Why mention this now?  Well, it's happened!  Within the last few days
(for me, anyway), what was a distant pipe-dream has become practical
reality, and Bob Billings said the same thing just a couple of days
ago.  What better tenth birthday note than to say "you succeeded!"
The progress made has been tremendous, it's been a global collaboration,
all in all a total vindication of the original aims of MMD.

The goal was to 'scan' a roll and interpret the results to reconstruct
the original perforation master, to replicate rolls punch-for-punch or
use the masters to create MIDI files.  It was spelled out most clearly
by Wayne Stahnke in the early days of MMD: I believe Wayne had already
developed his own process to do this and effectively challenged others
to do it as well.

This attracted a number of enthusiasts who have collaborated while
concentrating on whichever parts of the overall scheme most interest
them.  The detailed discussions eventually spun off into the
"rollscanners" group, so it is probably worth briefly mentioning here
what has been achieved.

The scanning itself became reality about three years ago, thanks
to Terry Smythe's enthusiasm and late introduction to the joys of
electronics construction, most of us have Richard Stibbons' "Mk3BT"
scanner board, but Gene Gerety's Rollscan-1 does the same thing and
showed how it should be done.  Spencer Chase lead the way in scanner
construction.  Many others have helped with parts, suggestions and
software.  The cost is trivial, the hardware really quite simple and
robust.

It's relatively easy to construct 'analogue' (okay, 'analog') files
from a scan.  It's much harder without a great deal of manual effort
to make 'digital' copies that reconstruct the original pattern of
punches on the roll.  This is the part that has recently been cracked.
A couple of years ago, Anthony Robinson produced a program that does
this under manual control.  Warren Trachtman suggested it could be done
fully automatically (using the same sort of maths as used for satellite
tracking, a "Kalman Filter").  His program has been improved over the
last year or so, and the version I received just over a week ago works
beautifully.  It's quite extraordinary.  You scan the roll, drop the
image into the software and within a few seconds you have a roll master
with no more than a handful of glitches, most reflecting perforating or
scanning errors.

The resulting perforation master is indeed held in a generic file
format -- none other than the global standard MIDI file (my 1997 post
got that bit quite wrong!).  The trick is to have one punch row
represented by one or more 'ticks'.  As it happens, Wayne Stahnke
posted a file of this form to MMD although he uses his own proprietary
format: his sample can still be found in the MMD Archives if you search
hard enough:

  Ampico 68283 Paderewski: Minuet, Op. 14, No. 1
  ftp://ftp.foxtail.com/pub/mechanical-music/misc/
  ftp://ftp.foxtail.com/pub/mechanical-music/misc/68283b.mid (86 kb)
  ftp://ftp.foxtail.com/pub/mechanical-music/misc/68283bar.mid (20 kb)

The art is to have an agreed way of storing all the other roll-related
data in the MIDI file, which is done in text events.

Coming on for three years ago I was lucky enough to be able to purchase
an 88-note Themodist / Duo-Art perforator made by the late Steve Cox
(Laguna Rolls).  I've been working in parallel with the others to
enable this perforator to produce proper one-to-one replica rolls,
including a perforation-level editor that meets all the picky needs of
UK Themodist roll perforation patterns and shows on-screen exactly what
you'll see on the roll.  Now, with Warren's software my in-house setup
is complete!  Bob Billings' report said he's achieved the same thing
(having built the perforator as well), and Dave Saul has been quietly
doing so for some time with use of Stahnke masters.  Only one part of
the challenge is left, to see the end of analogue roll copies.

This has been a major achievement all round, and MMD without doubt has
played a central role.  Happy tenth birthday!  What's the challenge for
the 20th?

Julian Dyer
London, England