Hi roll scanners in MMD,  if somebody would like my advice here it
comes.  (This only applies to discrete opto electronics, not CCD's).

There was not much success with reflected light so I use transmitted
light.  In 1990 I used an incandescent tube lamp because the roll paper
would glide softly over the glass.  The lamp was fed with about one-third
of the normal DC voltage.  I could see the nicely glowing surface of
the tube.  The infrared sensors saw the filament which is not at all a
straight wire.

There are matching pairs of Infrared Light Emitting Diodes and photo
transistors.  You can find them in a list under BPX81 and LD261.
They are so small that one hundred can easily be mounted in a row, 9 to
the inch.  But beforehand, measure them with an ohmmeter (9V battery).
You will throw away at least a dozen for "not good looking" and some
for being even too sensitive.

The LEDs can be fed via resistors from a regulated 5V power supply
(a power supply with an output that can be regulated).  The photo
transistors are in row with a trim pot (and a fixed resistor?).
The junction point is connected to a comparator (LM339) or simply
a TTL gate.  (I use a voltage-translating 12-to-5-volt gate here).
(Or vice versa?).

As an interface to the computer some 8244s will do, controlled by an
address decoder, but 8245s are much easier to mount (because of their
pin layout).

The distance of the opto electronics to the roll paper is each about
1/10".  Probably it could be more or less.  As LEDs and transistors
have an optical lens moulded on, the connecting light beam is very
thin.

You can adjust this environment until even a clear plastic will break
the infra red beams.  In practice the level will be higher, but, even
on translucent paper, the oil spots will not be read as holes any more!

Happy scanning !

Horst Mohr