In article , Paul Burridge
writes:

If one can vary the supply over a fairly wide range and the oscillator
only responds by very small changes in output frequency and doesn't
jerk into another frequency/output mode altogether, is this a
sufficient test on its own of that oscillator's likely stability in
the field?

Normally...but it may not satisfy everyone in here. :-)

What should be done in oscillator design is to make the minimum
gain through the feedback path - under all environment conditions,
including variation in supply rails (filament voltage if tubes) and even
component tolerances and tolerance changes - above unity by a
fraction. Then the opposite environment conditions should be
examined and the gain through the feedback path analyzed to see if
it exceeds unity to much. Too much gain might overdrive a quartz
crystal unit or generate harmonic garbage or whatever else might
be a no-no.

Once the above is satisfied, the prototype is then subjected to
actual environment changes, again everything from temperature,
shock, vibration, supply voltage, etc., etc., while measuring the
oscillation frequency changes with each environment condition.
To go really deep into that, one might get a number of active devices
with varying tested specifications and try them...but that would only
be for large production runs of the same thing.

If the frequency stays within the desired limits, it passes! Yay! :-)

In too much hobby work, someone grabs a schematic from some
publication and copies it, but usually with some parts changes ("for
convenience") and hopes it will work. Sometimes it does.
Sometimes not. :-)

Oscillators are actually a rather complex subject and take time to
get working within all the desired specifications. Before SPICE and
the ability to model ALL the various things in a circuit, all analysis
had to be done open-loop. The loop (for feedback) could only be
closed on the bench. That work was a real #$%^&!!!!

The best text I have on oscillators is a rare, little-known NTIS-
distribution contract report, AD 460 377, "Quartz Crystal Oscillator
Circuits Design Handbook," prepared by D. Firth at Magnavox
Company, Fort Wayne, Indiana, under contract to U.S. Army
Electronics Command, Fort Monmouth. It was obtained mail-order
in 1985 from the National Technical Information Service (NTIS) from
the U.S. Department of Commerce. Cost $8.20 for 478 pp, Xeroxed
on 8 1/2 x 11 inch paper. Original contract completion date is given
as 15 March 1965. [yes, transistors were around back then...:-) ]

"D. Firth" (name sounds very familiar) covered the whole gamut of
oscillator frequencies from 1 KHz (!) on up to 200 MHz, showed the
gain-impedance analysis of the feedback loop, and working units
done for actual environmental test and the test results which
included variations in not only supply voltage but also component
tolerances of each test unit! Marvelously complete report...can be
an example of how-to-do-it for anyone, including hobbyists who are
very fussy about their oscillator circuits, especially quartz crystals.
It's also a good tutorial that goes deep into circuit theory for
oscillators.

I'm NOT saying everyone ought to get that publication. There's lots
of good information around and a lot of that free (for quartz oscillators)
at the various quartz crystal makers' sites. Basic information on
oscillators, types, etc., is in lots of places and publications. The
final word on oscillator performance is its frequency stability over a
hobbyist-designated total environment.

Accuracy of a frequency measuring device to measure that oscillator
frequency is quite another thing. :-)

Len Anderson
retired (from regular hours) electronic engineer person