Sipke de Leeuw wrote:
>
> Hello all,
>
> Another small question for all the crystal experts: When I use a serial cut
> crystal instead of a parallel cut one, will the oscillating frequency be
> lower or higher than the frequency printed on the crystal?
>

I forget.. --TR

> Cheers:                                ... Sipke de Leeuw ...
>
> >>
> >> & more stuff deleted
> >
> >Well, it is on page 39 of the PIC16C8X data sheet I have.
> >
> >It should be a parallel cut crystal, as "use of a series cut crystal may give
> >a frequency out of the crystal manufacturers range".

Well, yeah, but just how much is that? I recall it being small; the
manufacturers specs on a decent crystal are really amazingly good until
you compare them to the inexorable march of time. The perspective is
that a five dollar watch is the most accurate instrument the average
person ever touches, unless they buy a ten dollar watch. The actual
tolerances of most electronic components are huge by comparison; a
microprocessor crystal is not intended to be a timepiece. With careful
design they can serve pretty well but you're not going to get what you
get in a cheap watch without proper design for function.

Its been years since I went through the exercise & I don't remember the
exact result, but around the time we were designing clock oscillators
with the first CMOS chips we did look at the difference that the
series/parallel switch makes in various circuit topologies. The
differences were trivial for our purposes. In the few cases in which we
wanted cheap watch accuracy we always designed the timing circuits with
the crystal manufacturers suggested layouts and drive circuits, and we
always included a trimcap. After about 1985 we usually used a 32768 Hz.
watch crystal in a carefully designed separate timing reference.

I have personally built some prototypes with PIC processors and watch
crystals that don't have worse than a one second error over several days
(in the lab); if I can do it, probably anyone else as stuborn out there
will prevail. I didn't do anything out of the ordinary that I can recall
except read my copy of the data sheet, which by the way doesn't mention
crystal calibration.

As a further perspective: Time Tech markets a device based on a 68000
that uses an ordinary 2.4576 MHz. crystal oscillated with the common
74HC14 circuit as a timebase. We oversample a ridiculously low level
signal synchronously with the mains frequency and convert the result to
an easily identifiable and repeatable physical parameter. If we tweak
the sync counter to provide 61 or 51 Hz. we see the expected 1 Hz.
interference as a (large) rise & fall of the measured physical quantity.
When it's set correctly we have never seen a drift over the course of an
experiment, which runs over several hours during testing. For real world
design I would expect anyone could get this sort of performance if they
are willing to put in the time identifying & considering all the
parameters.

And that's enough of that stuff.

-- TR