I can see two issues involved with a vibration test on an oscillator:
A) Changes that result in frequency drift - i.e. more of a after-the-test frequency shift. This is addressed by your requirement #1.
B) Phase noise of the oscillator as it is exposed to the vibration - more of a modulation of the output signal - usually referred to as microphonics. This seems to be what is addressed in your requirement #2.

The issues doing the vibration at several fixed frequencies Vs a sweep are the same issues you would encounter in a vibration test where you are looking for mechanical resonances. In this case, vibration of a particular component or flexure modes of the device may only occur at particular vibration frequencies, which a fixed frequency vibration test may fail to excite. It is changes in the capacitance/inductance/resistance or piezoelectric effect of components under vibration that cause microphonics, and the vibration is at it's maximum when you sweep across the mechanical resonance of a part.

Also see this previous thread on microphonics:
Thread247-78820

Comcokid,

        Thank you very much.  The random vibration test did show some resonances in the frequency range.  However, electrical performance (phase noise from 100Hz to 1 MHz and Single - Sideband AM noise) showed performance within specifications.  Since the vendor carried out the sine vibration test at particular frequency only (at resonance) I was wondering whether we were overlooking any other aspect of the electrical functional test.

        The reason for carrying out the sine vibration this way as stated by vendor "was being generating too much of data at other frequencies".  This may be true or they may not have proper instrumentation.

         I think we can accept the vendor's test as long as the sine vibration is carried out at resonance frequencies.
Thank you very much for the help.

Regards.

I would think that the requirement should tested exactly the same as it would for a resonance search.  

You would therefore sweep the entire frequency range of 20 to 120 with sufficient dwell at each frequency to determine if there are any problems.  A subset is a subset, so you run the risk of not detecting a problem.

If there are anomalies, you would treat them like you would resonance peaks and would then perform an endurance dwell to see what the problem is and to see if anything breaks.

TTFN

IRStuff,

        Point well taken.  This is on similar lines that we run pre and post random vibration sine sweeps to detect any failures during radom vibration test.

        In your opinion what would be the acceptable sweep rate for sine sweep particularly for microphonics considering vendor's view of "too much data"?.  I want this test being run properly without compromising the test objectives.  At the same time I do not want to go overboard.
I have noticed that our specification does not state the sweep rate for the sine vibration.

Thanks,

Depends on what you're trying to accomplish with this test.  If it's a design verification test, then you'd need to dwell long enough to find out if there are any problems at each frequency.  That could be seconds, minutes...

On the other hand, if this is part of acceptance testing, then sampling may be the way to go, particularly if, based on the first article design verification testing, you've identified specific frequencies that may give you problems.

For military systems, we generally perform sine sweep only during design verification and we use temp-vibe ESS to do product integrity testing as part of acceptance.

TTFN

IRStuff,

        Thanks,

        Our "Quality Conformance and Qualification Cross-Reference Matrix", show that requirement for microphonics verification for qualification as well as acceptance is by test.

         Since this is the first test on this OCXO, we will have to do the full sine sweep from 20 to 120 Hz as suggested by you.

         Thank you very much for the help.
Regards,