Robbie queries:
> [ Perhaps a strip of tiny music-wire or brass contacts could be fitted
> [ to the comb to make an electric connection as the tine is plucked.

Yup, I've tried that too.  I have a 55-tooth (including bells) 5-bell
wonder for which I actually _know_ one of the tunes.  Even have the sheet
music in two versions.  The problem this box has is that the comb was
very badly rusted when I purchased it by mail order many, many years ago.
One would think that, knowing the pitch of the bells and one of the
tunes, finding the tuning of the comb should be fairly easy.

I should mention at this point that there are absolutely no markings on
the comb to indicate how it may have been tuned.

I tried two methods to map the "known" tune.  The first was to physically
make a note of which teeth were plucked vs. the number of turns of the
fan.  (It took a long time, my back was killing me and I just about went
blind.)

 I transposed the results into Cakewalk, a MIDI sequencing program, with
the hope that by assigning probable notes to the various teeth and
knowing the tune the results would fall out on the floor.  No such luck.
I still have not been able to crack the combination.

Not completely daunted I decided that maybe part of the problem was the
accuracy of my recording method, so I set up a system much like Robbie
suggests.  It worked beautifully on paper, but in practice not at all
well.

The object was to record the point at which a tooth drops as this is when
the note sounds.  I set up simple contacts to connect directly to a
cleaned portion of each tooth (very tedious, by the way) and a bank of
LED's.  (I didn't want to get too sophisticated yet.)  I even incorporated
an  individual vertical adjustment for each contact.

The problems I ran into were many:

(1)  First off, direct contact to the comb proved quite unreliable, even
with everything cleaned up nicely.

(2) Secondly, the vibration of the comb once a tooth was plucked smeared
things more than a little.

(3) And thirdly, bent and missing cylinder pins contributed significantly
to the unreliability of the record.

Of the ideas presented so far I think the one suggesting a bank of micro
or optical switches operated separately from the comb shows the most
promise.  Again there are problems, but if one has the tools, patience
and mechanical ability they shouldn't be impossible to overcome.  Some
significant considerations that come to mind are the need for:

1. Some means of easily attaching the whole thing to the box in a stable
fashion.

2. Adjustable spacing, reliably variable to about .01", since not all
combs have the same tooth spacing, some being very fine (and hence a lot
of switches may be needed in a small space).

3. Easy setup and the ability to conveniently alter the pickup size to
fit large and small boxes with widely varying numbers of teeth.

4. If a MIDI interface is used some means of making it capable of
recognizing that a large number of teeth may be tuned to the same note.
This would, of course, require some signal processing in that the "tooth
drop point" of one note should not interfere with that of another.

5. If possible, to have it track the music in real time without running
into mechanical problems with the pickup.

So, folks, that's my 2 cents worth, offered in the hope that others
starting on the project will learn from my experience and that angels
will watch their step.

Lots of luck!

Jim Heyworth,
Sechelt, B.C., Canada