Hi, John!  I was pleased to hear from you, and of course I am happy to
provide you with a description of my roll scanning process.

I sense the presence or absence of holes in the roll pneumatically, using
pouches and microswitches, but no valves.   The paper is advanced
continuously at a speed of about 10 mm/s by a take-up spool.  The units along
the length of the roll are displacement, sensed by a measuring roller coupled
to an optical shaft encoder.  This arrangement captures the webbing, but no
visual effects.  The end result is a roll image in my own data format.  I
would be happy to provide to you with a description of the data format on
request.   The data represents the x-y location of each hole; all
perforations are treated in the same manner, with no attempt at
interpretation of the data.

It should be clear from this description that I have not broken any new
ground in the methodology or implementation of roll scanning.  What is new
about my work is its accuracy.  The location of each hole is determined
precisely enough to allow further processing, in particular the
reconstruction of the master roll.

Traditionally, we have all thought of a roll as having the holes positioned
in discrete locations across its width, but continuously along its length.
This is an error, of course.  The rolls were manufactured by punching a row
of holes with the paper at a standstill, followed by a discrete paper advance
before stopping the paper and repeating the process.  This advance, which I
call the "punch advance," is typically on the order of 1 mm.  It varies from
manufacturer to manufacturer, and was sometimes upgraded by a given
manufacturer, apparently to increase accuracy.

Thus, a rolls should properly be viewed as a bit map of an x-y matrix of
pixels, each "pixel" consisting of the presence or absence of a hole.  The
fact that the pixels are larger than the punch advance, yielding a slot with
scalloped edges, is not material to this discussion.

With this picture in mind, it is clear the the goal of scanning a roll is to
reconstruct the x-y matrix and the (on-off) value of each pixel.  Stated
differently, our goal is to restore the master roll from which the production
roll was made.  When we have done this, we have scanned the roll exactly.

To illustrate this point, I have attached the file _68283B.MID to this
message.  The file consists of the recreated master roll of Ampico recording
68283, Paderewski's Menuet, Op. 14, No. 1, played by Sergei Rachmaninoff.  I
chose this roll because it apparently sold in large quantities, and is
available in almost every large Ampico roll collection.  I suggest that you
take the time to pull the file up on your computer screen and examine it
carefully with a copy of the roll (and a straightedge) at hand.  You will
find that all of the perforations are at their correct places, as determined
by counting punch steps.  A new roll made from this file would be identical
to a roll made in the 1920s.

Thus, my contribution to roll scanning technology has not been in developing
a new method or apparatus.  Rather I have contributed two things: (1) Setting
as a goal exact, rather than approximate, scanning of the roll; (2)
Demonstrating that this goal can be achieved.

There has been a lot of discussion lately about scanning of roll collections
on a large scale.  To my knowledge, no one has yet asked the question, "How
good will the scans be?"  I would like to suggest that we should consider
anything less than exact scanning to be inadequate.  We have been entrusted
with a precious legacy, and we have a duty to preserve and protect it.
Preservation of the performances on these rolls must not be corrupted by
scanner errors, or we will have failed to live up to our responsibility to
the artists and engineers who created them, and to future generations of
listeners.

With best regards, I remain

Wayne Stahnke

 [ Editor's Note: I have compressed the MIDI file (68283B.MID) with "zip"
 [ and named it 68283B.ZIP.  I have further UUENCODED that file and have
 [ enclosed it at the end of the digest.  Those of you with MIME-compliant
 [ mailers can probably open it as an attachment.
 [
 [ As Robbie explained it to me, the MIDI file is really an image of a
 [ Master Roll, intended to control a perforator, and therefore the timing
 [ is different from a MIDI file to control a synthesizer or a solenoid
 [ piano -- "the slots in the master roll are "lengthened", so to speak!
 [
 [ Enjoy
 [
 [ Jody