It sounds as though Albert de Boer has tried this!  In the old
pneumatic readers which read masters pneumatically, they were stepped
through together with the perforators.  It's the stepping process that
"syncs" the copy with the master, anyway, and it is this sync that was
largely responsible for good resolution by preventing holes being
punched during a paper advance.  So even though a line was still
"reading," it would not be punched, and if the line stopped reading
a bit early, it would still be punched before the advance.

That same stepping process can be simulated a variety of ways today,
and using pneumatic readers, at that.  Sometimes we tend to think in
terms of "high speed" readers going the roll tempo.  That cuts down on
the accuracy.  And without discrete stepping to the same stops and
starts or even finer, that roll will not be as accurate as it could be.

By using a good lens, you are able to count the steps/inch that a roll
is cut to by counting the number of nibbles/inch at the edges of long
holes.  By setting a stepper motor or a rolling counter of some kind
(like a mouse) on the paper, you can then control a variety of methods
to either blank or arm the reader that makes the file.  Think of it as
a "strobe." It will also keep track of the exact position of each hole.
I would think that principle there would be responsible to control the
reliability and overall accuracy of the file.

A pneumatic reader however (especially without any stepping control),
tends to "overpunch" the same way a light-operated reader does, and for
the same reasons.

The beginning of each hole is easier to sense, but all things have a
natural hysteresis that, once turned on, want to stay "on." I have seen
pneumatic switches that were not designed right for their job.  They
are either too slow, or too unreliable from actuation to actuation to
be trusted in such a job.

Electrical engineers trying to use pneumatics sometimes oversimplify
that system interface, and think they know how to make a pneumatic
valve work as well as anybody else, but, as it is turning out, they
have a few things to learn, too (for example, the load affects the
characteristics.  The less load you have, the faster it actuates and
the slower, percentage-wise, is the return.  So the hysteresis of a
pneumatic pouch or valve widens considerably when the only load is a
bleed).

But in any kind of roll reader, each system has its problems.  With
scanners, a problem might be roll tracking and very dark words on the
roll, tears in the edges, different kinds of tape repairs, and failure
to keep track of the roll position in time, accurately, and optical
resolution and distortion.  Light sensors in line have the "over-
punching" problem, tending either to cut off a hole when it still has
1/3rd of a dia. to go, or extending a hole for up to perhaps 2/3rds of
a diameter or more.  They also are very temperature intolerable, and
unless stabilized pretty well, become less and less reliable as they
warm up -- always unevenly, I might add.

The bubble sensor I suggested would tend to overpunch too, without
the other controls, only its sensitivity would still be more precisely
fixed than you could maintain with photoelectrics alone, so its char-
acteristics would be that of more exactly blowing up the size of each
hole -- magnifying it first before you read it.  Lint and paper dust
would not be its problem because its sensitivity would allow it to play
through dirt, and you can stuff some felt wadding in each line to
become a permanent filter.  Dirt would pack its way into the filter,
but would have no other effect.  (There are some nice things about
pneumatics, when designed for the job they are expected to do.)

Pneumatic readers, while slower inherently, have fewer problems to
solve, and when you know how fast and clean a GOOD pneumatic valve
specifically designed for the duty it must do can be, then you will
realize that the tolerances required to read a roll accurately that way
are well within its range, all the while removing the many other kinds
of problems you must face with optical, heated wire, and light sensors.

The Standard Pneumatic Company years ago measured the maximum speed of
a outside/inside valve combination operating (oscillating) on a bench
setup as, I believe, about 1250 cycles/second.  That was a technical
answer (claim) to those at the time who were saying that dual valves
were not as quick as single valve actions.  Taking that another step it
is possible to build a reader whose valves oscillate quiescently, all
together, synchronously, and a hole being read merely operates a cutout
pouch that transfers those pulses to the reader system.  The reader
system lines are also armed synchronously by the same "pulses" but
unaffected until the cutout pouch opens, completing the air circuit.
The valves are synched to  the speed of the roll transport, and become
the "strobe" as well as the sensors.

The bottom line is this: There is more than one way to skin a catfish!
What one person finds a solution for, another claims is impossible.
When I got into pianos, I listened to technicians saying and writing in
the Journal, "There is only one way to do thus and so, and if you don't
do it this way, it will be wrong."  Over the years I learned that they
are wrong.  And it was a very good thing, from time to time, to know
several ways of doing the same thing.  It gives you more appreciation
of the kaleidoscope of human talent, insight, and abilities.

Craig Brougher