I had a chance to read through some of the MMD Archives!  A thread
in the first week of March 1996, "Exact Scanning of Rolls", caught my
attention.  I wanted to see if I could decode the zipped and encoded
attachments from several of the posts.

 [ The thread begins roughly in MMD 960305; see
 [ the early articles by Wayne Stahnke listed at
 [ http://mmd.foxtail.com/archives/Authors/Aut318.html
 [ and also a later article, with sample files, at the MMD
 [ Tech site: http://mmd.foxtail.com/Tech/transcribing.html
 [ -- Robbie

A few years back, when these encoding formats were becoming popular
and starting to show up in email, I wrote some PostScript programs for
decoding base64 and uue encoded files.  I still sometimes use these to
verify that a binary is what it portends to be.

I decoded the Ampico 68283 roll files and found I had some data to play
with.  I modified my PostScript MIDI analyzer and ran it on the data.
Now the problem I had before, was that I could read the MIDI and show
the display list.  I was unsure how to display it graphically.

The file "68283b.mid" gave me the clue.  Wayne Stahnke encoded this
data as one punch step per MIDI tick.  Now I could figure out how to
display the roll.  130 points is a bit less than 2 inches of paper.  I
figured I could fit 4 strips per page.  By repeating the array which
I called "Bar" I could duplicate the action of the punch.  A note-on
event would be stored in the array.  When I got a note-off event I
would step the array across the section of the page.  Then I would
update the Bar array.

What a thrill when I ran this and saw the roll number as punched
into the tail of the image.  I had on the screen what looked like a
miniature piano roll, with extra stuff on the sides.  [The "notes"
at the edges are Ampico logic control channels.  -- Robbie ]

I then tried to read the 68283BAR.mid file.  This looked like modern
art, with the holes really elongated.  I could see the roll number
really stretched at the end.  I tried this file at home on the
Macintosh Midiography program.  It played like a dirge with bells.
(I did not know what to expect.)

The file 68283BV1.mid Seemed to be like 68283BAR.mid, but encoded with
note-off events and the expression in channel 2.  This file seemed to
be a playable file.  As I was using GM [General MIDI] under QuickTime
it sounded like a harpsichord.

As for the original thread, I can answer the question asked.  Yes,
I was able to read the MIDI data nearly 5 years in the future.  I even
had less trouble, because decoding programs, like WinZip, are more
robust now than they were then.  I even was able to take the data
cross-platform.

Later postings indicate a suffix of WEB.  This is actually confusing.
I take it from Wayne's comments, that this is similar to the XXXXXb.mid
file.  I really like having the raw 'punch list' to work with.  It is
much easier to go from that to other arrangements, than is to go from
a MIDI interpretation and try and quantize the data.

 [ The WEB file controls the perforator; that's why all the long
 [ notes chatter like a woodpecker!  The BAR format file is more
 [ like the data sent to all the valves in the pneumatic piano.
 [ -- Robbie

This quantization, has given me trouble in the past when I have
tried to enter music note by note.  The composing software, a really
old program called Concertware, always tries to even out the measures.
As I am attempting to input 'clockwork' music I find that the music can
be arranged in-between.  What was really giving me trouble was a
misunderstanding of velocity.  When I input triplets or other grace
notes, these would blur together.

Julie Porter

 [ The key feature of the WEB file is that it controls the perforator
 [ _exactly_, punch for punch.  Richard Tonnesen and Dave Saul use this
 [ format to assure that any and all production rolls are absolutely
 [ identical.  But WEB format is too detailed for routine music
 [ composition, and so Wayne has provided some utility routines to
 [ convert the special WEB and BAR files into MIDI file format.  (MMDer
 [ Richard Brandle has similar capabilities in his powerful music roll
 [ editor, "Wind".)
 [
 [ Since your interests lie outside piano music (which is what normal
 [ MIDI files are best suited for), and include automatons, I believe you
 [ will find that Wayne's "View" editor provides the absolute control you
 [ need.  (Plus it can handle 240 channels, if you someday make a monster
 [ automaton!  ;)
 [
 [ -- Robbie