The simplistic approach of utilizing a single 'inverter'
stage to be used as a crystal oscillator has always
been bad design practice in the past and is always
fraught with it's own set of perils.

A little practical History.

I re-designed a 10.24 MHz oscillator years ago that
used a TTL gate. The original oscillator and crystal
were overly subject to frequency changes with
temperature. Temp compensation was attempted
on the crystal and only partial success was had.

A discrete 2N2222 design was drawn up, built and
the same crystal now showed 10.245 MHz in the
new circuit! This new circuit with the original crystal
was less responsive to temp change than the TTL
gate design leading to the conclusion that *inherent
circuit propagation delays* in the TTL gate over
temperature were responsible for the excessive
frequency change with temp. This lead to speculation
that the inherent properties of the TTL gate contributed
substantially to frequency instability versus temp and
the prop delay also contributed substantially to the
off-frequency operation (the proper freq of operation
of the cystal was actually 10.245 MHz but ran at
10.24 MHz using  the TTL gate). I also seem to
remember a problem with starting or 'squegging'
(improper freq of operation) of the TTL based
oscillator.

Unfortunately, the new design with the same circuit
was oscillating 5 KHz higher in frequency. Padding
a 10.24 MHz crystal down 5 KHz is not done (this
is/would be 'excessive warping').

A new crystal was found and all were pleased in
Mudville with the 2N2222 based 10.240 MHz osc ...

I don't know what MChip (or others) actually has
fabricated on the die as an 'oscillator' stage but it
probably isn't a well designed analog 'oscillator'
stage with suitable feedback or biasing.

The moral of the story: It's "potluck" with the
onboard oscillator and an external crystal all of
unknown specification.

Jim P


Dan wrote:

<snip>

I understand problems with Scenix, as they are running 50 Mhz, which
happens to be in the "critical" frequency range where RF effects start
to take over, ie transmission/reflections/harmonis/etc, and where
fundamental mode crystals are hard to come by, but 16 Mhz with a PIC?

Any considerations as to the real problem here ???

1. Crappy Mchp engineering.

2. Crystals and resonators today are just too different from each
other, due to a profusion of new xtal manufacturers, so every design
situation is unique.

3. Too simplistic design approach by users. We have come to a point in
time where we really need Smith Charts, Q and loop gain measurements,
transmission line calcs, etc/etc, just to duplicate what is easy to
do on other controllers - even at relatively low frequencies.

4. Physics has changed in the new millenium.

5. Something else [phase of the moon, karma, prayer, etc].

Cheers,
- Dan Michaels
==============