As a related project [to rebuilding the Ampico B] I decided to build
a roll reader to preserve my collection before the paper started to
crumble.  A secondary goal was to accomplish this without major expense
or specialized machine shop facilities.  I believe that this was
achieved.

The scanner is built in a plywood box about 18 inches on a side, with
a pyramidal top to hold the scanner.  Conventional roll chucks hold the
supply spool.  The paper is drawn over a reading bar of black Plexiglas,
with a slot cut at the reading position.  This has a black-lined cavity
behind it to act as a light trap and insure a good black background even
with quite translucent paper.  The paper is taken up on a 6-inch dia-
meter drum, driven by a stepper motor controlled by a variable oscilla-
tor to set speed.  The large drum gives more gentle handling of failing
rolls, and keeps the scan speed more steady.

The scanned area is lit by a row of small incandescent lamps in a
reflector of elliptical cross-section.  This gives a highly diffuse
uniform light to maximize contrast with various papers.  The light
is controlled by a photocell feedback arrangement to insure constant
apparent brightness of the paper, independent of its aging or degree
of transparency.  The paper motion is sensed with a "retired" PC mouse
located on the reading bar out of the optical path.  This signals the
electronic components such as the scanner to set scan timing.
Currently I am scanning at 200 scans per inch of roll motion.

The scanner is one of those hand-powered ones that came out several
years ago, and have virtually disappeared since then.  It was designed
to scan a 4-inch-wide path at 400 pixels/inch resolution.  I mounted the
scanner about 10 inches above the paper path to increase the width to
about 13 inches or so.  This threw the image badly out of focus, so a
visit to the local discount optician found a hard contact lens that
nicely fit the scanner lens barrel and corrected the focal length to be
within the adjustment range.  Being nearsighted myself, figured the
machine could use the same type of fix.  Final adjustment of scanner
position gave a resolution across the paper of 108 points per inch, and
a span of at least 13 inches without distortion.  The light source and
timing roller in the scanner were disabled.

That about covers the scan system, most of the rest is software.  The
scanner interface to the computer is limited to about 200 scans per
second, or a "Tempo" of 50, which is tolerable.  The limit is in the
hardware, and it's not worth trying to speed it up.  At that speed, I
can save the image to disk, provide a full-screen display of the
scrolling image as if watching the paper, and the computer is loaded to
about 30% of capacity.  If I see a problem in the image, a mouse click
on it stores its location in a log file for editing later.  For this, I
haven't found an image editor that will handle an image 1400 pixels
wide by up to 400,000 long, so I made utilities that cut the file in
2-foot pieces, allow editing each as needed, and put them together again.

Another software program first analyzes the image, precisely locates
the track centers, determines the punch advance, and for Ampico B
rolls, measures any skew of the punches relative to the paper path.  No
tracking is done on the scanner, as it is easy to do it on the image,
and can better ignore damaged edges.  A separate program uses these
details to generate a data file in Wayne Stahnke's *.web format.
Others can convert these to *.bar files, or *.mid if desired.

I have scanned about 300 of my Ampico rolls, along with a couple dozen
each of Duo-Art and Welte-Mignon rolls, and 40 or so 88-noters.  The
design can handle any roll up to 13 inches width, such as an Estey
Residence Organ roll I also read.  Work continues on polishing the
software for better accuracy, and "smarts" to ignore image errors
without manual editing.

Bill Earnest
Allentown, PA, USA
wde@fast.net

 [ I worked today on the links at the MMD site and also created a
 [ page for roll transcribing activities.  Wayne Stahnke's package
 [ of tools for editing and evaluating transcribed music rolls,
 [ including his transcription of the Rachmaninoff Prelude, is at
 [
 [     http://mmd.foxtail.com/Tech/
 [
 [ The links from the Gallery page should be evident.
 [ -- Robbie