I thought that Horst Mohr's idea of roll scanning was really a good
one because it takes out the biggest problem -- selecting the right
phototransistor pair, and because it eliminates the main problem with
roll reading -- dirty paper.

I want to do this someday, but not right away.  Still, every idea is
welcome to me, and I suppose to others interested as well, so if I used
LEDs and phototransistors, I'd do it Horst's way, with an additional
twist.  Whatever I came up with though, I would want a test like this:

I would want a test roll made up of nothing but chain-bridged holes
with stops and starts at every percentage of the increment, all the
way across the note sheet.  When I was ready for a final test, I would
play that test roll on the reader and electrically determine if any
reader missed as much as a single hole in the entirety of the sheet --
at a normal roll speed of about Tempo 70 if I planned to scan at a
Tempo of 20 or 25.

Granted, the reader isn't going to operate that fast, ever, but a test
should be made nevertheless.  I would then scroll through the computer
looking for a "hole." I think that the reader should be able to read
chain bridging perfectly  at full speed without a miss before it is
able to read the lock and cancel holes of an Ampico roll well.

(I worry a bit about the reaction and recuperation times of some
roll-reading sensors, as well as accumulated error with time, dirt,
fuzz, temperature, etc.  Because when you want to make the reader pay
for itself by doing lots of work and cutting your own efforts down, you
have to make things easy and slick and trouble-free.  You have to be
able to run it continuously for months at a time without having to tear
it down to clean it.  Sensitivity is not as much the issue as response
time, or slew rate, in my mind.  And if it takes a lot of cleaning,
that really cuts down your work output.)

One advantage of the phototransistor pair Horst cited was the thin,
focused beam it is able to read.  When you try reading a light bulb,
the light is un-collimated, allowing the sensor to trigger at the most
extreme angles to it.  So it doesn't give you much control over the
sensitivity or the slew rate.  These focused LEDs mentioned, _coupled
with synchronization_, will give you a very accurate reader.

If your reader were to use a stepper motor (surplus) you will have the
best of both worlds.  You can buy these and a motor stepping driver.
That means, your driver turns the armature of the motor in increments.
You can find the exact number of steps/foot a number of ways, by exam-
ining your roll, of course.  Set your stepper to that increment times
a multiplier (a minimum of 3).  You can do this with a combination gear
drive and a control on the driver, or 3 different diameter steel
rollers on the paper (Drive with steel, clamp with rubber).  Having
three gear ratios this way should give you about any increment you
desire.

If I were to build a transport, it would drive the paper, rather than
the take-up spool.  The take-up spool then would just be clutched to
get rid of what was coming out of the reader.  The paper would be
surface- driven, pulled across the reading head against a play brake on
the roll chuck.  Frayed edges would  be edited out.  This scheme would
factor out paper buildup completely, and there would be no reason for
tempo variations from the original roll.  It would also factor out
paper stretch and slippage on the take-up spool.

On the driven roll, position a counting wheel read optically.  This
will become the sync pulse for everything by turning the LEDs off and
on, or the output lines off and on, or whatever is convenient and
spikeless to sync the entire reading operation to the final roll speed.
If things are not synchronized, it really doesn't matter too much what
else you have done, because the unit will just free-run and accrue
errors in tempo, etc.  With sync you have control, and can use it also
to edit.  Also with sync you can use it in a troubleshooting program to
check out your equipment.  Without it, you are just shooting in the
dark.

The main reason for sync as I see it, is that cut-off times for holes
should be almost perfect, since the resolution is going to be at least
three times that of the read roll, sync-wise.  The reason for finely
collimated light beams is equally important.  The thinner the beam,
the more latitude you have and less need for mechanical precision in
the roll reader.

If you visualize a mechanical roll reader that used steel pins which
have to drop through a hole, you would understand why sync is
important.  If the paper had tractor drive to keep it in perfect sync
then reading pins could drop into and out of the perforations in that
paper with perfect precision, because of the precise stepping.  It
would be like a high speed pin printer whose lines of print are very
precise.

Some of those old pin printers could go so fast that they could
actually throw a stream of paper clear across a small room before it
ever hit the floor.  Each line of type was exactly the same as the
next.  Without the stepping, however, even the best roll-fed paper
under a pin reader cannot read half of the beginning/ending perfora-
tions.  They will catch the ones in-between, of course.  But to be
precisely at the beginning and precisely at the ending of a perforation
would be almost impossible without sync.  So if the pin doesn't drop
(and it might miss by only a few thousandths) you don't read it.

That is exactly what happens with electronic readers, too.  They are
slow to recoup, like a steel pin is, so they elongate some holes
terribly, unless you read them at a snail's pace.  Likewise, they vary
in their ability to turn on as quickly as their neighbors, so you also
see a variance at the beginning of a hole.  If the machine stopped for
the pin (or the light beam) like a bus picking up a passenger.
Everything has time to settle in and read synchronously, and nothing
can be read until the circuit is rearmed at each step of the way.
That's why sync is very necessary.

Finally, the stepped pulse can be divided down and used to trigger
the step function in Cakewalk if you wish.  Each time a step occurs,
Cakewalk looks around to see what's happening.  If there's a perfor-
ation, it creates it.  If not, it leaves that increment blank and
goes on the next increment.  You will step yourself right through the
program with your sync pulse and everything will be synchronized with
the measure lines.  I don't know if Richard Brandle's Wind program uses
syncing, or not, but that would be a nice feature to have.

Craig Brougher