While an optical line scanner and software analysis will go a long way
to recreate a roll master, there is a technique which can eliminate the
hole diameter from the data and allow the decoding to concentrate on
the apparent center points of the punched holes.

If you remember the illustrations which describe solar and lunar
eclipses, you will remember that the moon casts a large penumbra
shadow and only a pinpoint umbra shadow on the earth.  Inverting this
technique, because we want to cast a shadow of a hole rather than
a disk, we use a ring light source.  It will cast shadow of a hole
as a large dim ring with a bright dot in the center where the light
from each position of the ring comes over the edge of the hole and hits
at the center of the hole some slight distance away.  Perhaps this
distance can be set by the thickness of a sheet of glass.

If you put high contrast photographic paper, ozalid paper, a photo-
transistor sensor bar and set a threshold or focus a sensor at
that focused dot level, you will get a very concise event for each
hole.  Where the perforations are scalloped, you will get stronger
light at the centers but a generally bright line all along.

Now, if you want to get only a bright light at the centers of all holes
_and_ the perforations, use only partial arcs at each side of your
light ring.  These will combine to create a bright spot only at the
center as the light drops over the curve of the scallops.

Simplifying further, mask the roll except for a tracker hole width,
strip down the playing zone (or use a line reader) and use only two
lights, collimated to provide parallel light (or very far away).  The
only combined light from the two will be where the holes are full and
the scallops are full.

I appreciate that this may well be academic, and sending rolls and
money to Wayne Stahnke (if he will have them) continues to be my idea
of perfection.

Karl Petersen
Washington, Illinois

 [ Good idea!  This is an application of optics which nicely matches
 [ the light source to the desired property of the object.  You
 [ might even call it "source impedance matching'!  -- Robbie