Ok, I just ripped apart the 'buzzer' and it is not a piezio
element at all-- it is electromechanical.  (Two wires to the box,
inside the box is a small signal transistor, a couple diodes, a
fat coil, and a spring.)  No wonder it wasn't working well.

        I guess I'll pick up a real piezio buzzer and hope it works
better.  As I want the thing for debugging, it would be very nice to be
able to hear 50ms-100ms blips, like someone suggested.  (I do not want
to use interrupts to drive a debug device, especially since TMR0 is
already used.)  So I'm not much into the idea of having the target
generate a carrier and modulation.  (That's a job for a 12c508..
anyone want to build a quick & dirty audible logic probe?)  1-2Hz
modulation seems a little slow, especially for debugging a state
machine with 8-16 states!  That's a lot of time to sit there counting
buzzes.

        It beats an LED for debugging!

> > So I stuck a piezio buzzer on PORTA.  It works as a good in-circuit
> > debug, but it is rather high overhead (ie: set the port pin, delay for
> > 400ms, clear the port pin, delay for 400ms, repeat for more buzzes).
>
>   It certainly does sound as if he is using a piezo "sounder" with an
> internal self-excited oscillator, by which I mean it is a three-terminal
> disc providing feedback to a (single) transistor.
>
> >  The buzzer seems to have some odd characteristics also (I don't know
> > much about piezo buzzers-- startup time seems a little long, etc).
>
>   This characteristic is *typical* of self-excited oscillators with a
> very high "Q" of the disc.  OTOH, they are *very* efficient (noisy).
> The "magnetic" type of buzzer (coil; magnet fixed to reed; plastic
> diaphragm) using the same principle is also very slow to excite, maybe
> just a little faster if you feed it plenty of power (and they consume
> *much* more current).
>
>   For these reasons, these devices are quite poor for Morse code.  A
> separate electronic oscillator, particularly of the "R-C" or
> "relaxation" type rather than the various "tuned curcuit" (State
> Variable, Wien, Twin-T) can be started much quicker, or a continuously-
> running oscilaltor can be gated to your transducer.
>
>   But this isn't what was wanted.  Point is, you get the PIC to generate
> both the "carrier", say 2kHz and gate it with the "modulation", say
> 1.25 Hz as suggested, and feed that to the transducer.  Unless you are
> very lucky (or tune the latter!), it won't match the piezo disc
> resonance, but it certainly will go on and off smartly.
>
>   As such, it (or at least the continuous-wave 2kHz version) is a far
> better diagnostic than a LED because the ear is *far* better at picking
> interruptions and characterising brief transients than the eye.
>
>   On another list I mentioned the virtues of an old-fashioned "crystal"
> earpiece as diagnostic tool.  I'll try and bring that up on my webpages
> sometime!
>
>   Cheers,
>         Paul B.
>