>> Thanks for all of the help.  I guess my first step is to go get another
> crystal
>> that is definitely designed for parallel resonance.  How do you tell if a
>> crystal is designed as such?  I looked in my Digi-Key catalog and they don't
>> have anything called a parallel crystal.  Can you assume that if the load
>> capacitance is given, that it is parallel?
>
>Yes, this is it exactly.  That load capacitance figure is how you
>determine the value of the C's in the xtal ckt.  These C's are there
>for the xtal, not the pic, so use the xtal specs - not the PIC data
>book.

I think the capacitance rating is the total effective capacitance expected
to be in parallel with the crystal (the two load capacitors in series plus
stray capacitance inside the PIC plus or minus various nonlinear effects).
Using other than the recommended capacitnace causes the frequency to change
slightly, an effect that is usually masked by the imprecision of cheap
crystals.  Using a "series" crystal in a parallel circuit like the PIC
oscillator will still give stable operation, but at a frequency far from the
crystal's rating.
>
>> For the price of some of these
>> crystals, I can get a TTL clock oscillator.
>
>Crystals are typically about $1 in low qty - more for lower
>frequencies (~<3MHz)

The low-cost crystals (and oscillators) often have a tolerance of +-50 or
+-100 ppm (0.01 % in the later case).  This is great for most applications,
but certainly not the limit attainable with quartz crystal technology.  More
precise ones can cost $5 to $10 per unit.  I take it that a timing error of
several percent is involved here, so this isn't a concern.
>
>In the PIC data book, they show
>> a buffer on the output of the oscillator, is this required if I am only
>> running one device (the PIC) from the clock?
>
>Generally no.  Reconsider if the oscillator is located far from the
>pic(?).

Make sure the Voh of the oscillator is high enough.  Many of them put out
only 3V or so, which is fine for TTL work but doesn't meet the PIC's
specification of 80% of Vdd (as I remember).  Often a pull-up resistor will
cure that, if not look for an oscillator with a CMOS rating, or use a buffer
with CMOS-compatible output.

Also, when investigating clock speed/stability problems, probing the X1 and
X2 pins is generally going to disrupt the circuit.  Instead, load a PIC with
a test program that toggles a pin as rapidly as possible and monitor that
pin (this also makes sure the problem is really the clock rate and not the
software).

-Mike