> From: "David E. Queen" <dqueen@nnl.net>
>
> I got my interface test unit built and running.
>
> To my surprise the times were not as expected. 241, 482, and 964 milliseconds
> not 250, 500, and 1sec. I had double checked in MPLAB and I am using the
> right number of cycles. My problem is my crystal is 4% high in frequency.
> About 4.16 to 4.18 depending on where I try to measure it.

4% is awfully bad for an xtal.  They are usually spec'd to within 100ppm
(0.01%) if correct loading is used.  Try a different xtal.  Note that there
are 2 basic types of xtal: series and parallel resonant.  I can't remember
which, but one should be in an inverting loop and the other in a non-inverting
loop.  Don't know which type the PIC is expecting.

>
> I am using a 16c84 rated at 10mhz, a 4.000mhz crystal and 2 22pf caps from
> Digikey. I set the oscillator bits to HS. The unit was built on one of MEL's
> proto boards.
>
> What are the key factors I need to look at when I need the oscillator to be
> within 0.1% without tweaking the frequency?

If used correctly, the xtal will be within 0.01%.  Try adding a series resistor
since the xtal may be being grossly overdriven.  Perhaps the seal is broken
(although I would expect moisture on the quartz would slow it down not
speed it up as you state).

>
> I really thought I could just grab a 4mhz rock and a couple caps ant it would
> be at least as close as I need. Should I consider using a canned dip
 oscillator?
>

That would certainly work.

>
> Another question is can you use a solder less breadboard (I just got one of
 the
> E.N.D. unit's) and a 20mhz clock on a 16c65 windowed part. I have concerns
 that
> the oscillator will go nuts with all the stray coupling, and I will need 0.1%
> accuracy on that unit too.
>

I have had no problems with 4MHz logic (was a Z80 system) done up on a
breadboard.  Using a frequency meter and changing loading capacitance by
a few 10's of pF produced only a few 100Hz difference at 4MHz.

Regards,
SJH
Canberra, Australia