A description follows of the roll reader that I intend to build.  I would
appreciate any comments and suggestions that might help this project
along.

The reader will initially be built with a reader head for Duo-Art and
88-note rolls.  It will be built to accommodate wider paper and to allow
reader heads for other formats to be substituted easily.  The reader head
will consist of .010-inch acrylic fiber optic threads sandwiched between
Teflon bars.  This provides a smooth continuous surface that is virtually
immune to the effects of dust, etc.  Individual fibers can be sorted and
bundled to represent hole positions, or channels can be cut in the hole
positions only and filled with fibers.  The collected fibers will impinge
on individual phototransistors, one per hole position.

The phototransistors can stay in place in a rack and various scan heads
can be substituted by having their fiber bundles held in a matching rack.
I have experimented with this approach and it seems to be able to detect
minute paper displacements.  Because the sensitivity is very high,
detecting the onset of round holes should not be a problem even with minor
tracking errors.  I plan to experiment with detector window width to
determine the optimal size for minimizing tracking error and adjacent
hole reading.

The output of the transistors will be connected to an array of 8- or
16-input analog multiplexers depending on whether an 8-bit or 16-bit word
size is chosen.  The outputs would be fed to an array of comparators with
variable threshold to allow for varying opacities of paper.  (This could
be dynamically adjusted through computer control of a D-to-A converter.)
This would produce either 6 16-bit words per scan line or 12 8-bit words
per scan line.  This data would be fed to a digital input-output (I-O)
board in a PC.  The multiplex addressing and paper stepping would be
controlled by the same board.  I believe this approach would be quite fast
as far as the hardware is concerned.  Scanning 200 steps per inch should
not be a problem.

I have several concerns regarding the software to control the data
acquisition.  I do not have experience with this and would appreciate the
opinions of those who do.  If a high level computer language were used to
control the I-O board and write the data to a file, would this slow down
the scanning to a crawl?  Robert Hopp suggests that he collects data and
converts it to MIDI format and plays it simultaneously.  I would love to
do this too, but I don't want to spend my life writing assembly code to do
it.  If rolls are to be scanned and archived, it is important to have
confirmation that the collected data is accurate.  Does anyone (Robert
Hopp?) have software that they would like to share that does this.

I am willing to devote the time to develop an easy-to-build, reliable,
and not very expensive scanner that can be used by a number of people to
help preserve a musical heritage that is rapidly falling to dust.  My
only reluctance is also having to develop the software, especially if it
has already been done by someone who is willing to share their efforts.

Spencer Chase

 [ Editor's note:
 [
 [ Our member Horst Mohr built a phototransistor reader with stepper
 [ motors, similar to your description, Spencer, and I hope that you and
 [ Horst can discuss your project together.
 [
 [ By the way, I substituted "reader" for "scanner" in your article,
 [ because the term "reader" has been associated for decades with parallel
 [ data devices such as punched paper tape and punched cards.
 [
 [ The term "scanner", in the parlance of industry, is used mostly
 [ for devices which move a single element across the object field,
 [ for example, (a) the cathode-ray scanning beam of a TV or video
 [ monitor, and (b) the single-point scanner of Robert Hopp's "Amadeus".
 [
 [ The "parallel reader head", seen in both player pianos and Hollerith
 [ machines, is used to achieve the greatest possible speed (e.g., frames
 [ per second while scanning music rolls).
 [
 [ Robbie Rhodes