[time-nuts] Frequency Stability of Trimble Mini-T
SAIDJACK at aol.com
SAIDJACK at aol.com
Thu Oct 16 16:40:07 UTC 2008
Hello Mike, Tom,
just came across your post. This looks to me like a classical crystal
frequency jump that the unit is compensating for. Looks like the Mini-t is a bit
under-damped in it's response.
John Vig talks about these jumps in his paper, and he mentions that the root
causes are not well understood. One way to verify this is to plot the EFC
voltage (not sure if this can be done on the Mini-t), and see if it "jumps",
similar to the plots in the Vig paper.
Co-incidentally we have developed a unit for this same Army requirement
(11Hz requirement), and early on had seen a similar issue which we solved, and we
can now easily meet this requirement. Let me know if you would like more
details, we can provide these offline.
Typically crystals jump <1ppb, but I have seen worse. That busts the 11Hz
requirement of course. With GPS disciplining, these jumps can now easily be
seen. "Curing" the jumps is a combination of crystal blank processing, testing,
burn-in, and other factors. There is no easy solution.
You actually have stumbled upon an area where, as well documented as Quartz
oscillators are, there is quite a lack of documentation and research in my
One note: some of the time nuts had received E1938A' units that had tags on
them saying "crystal jumps", so this problem happens in the best families :)
In a message dated 10/14/2008 17:13:04 Pacific Daylight Time,
mfeher at eozinc.com writes:
Thanks for your reply. I was not present during the testing. In my opinion,
and now I have been able to convince others, the contractor passed the test,
which is derived from MIL-STD-118-164ACN2. The main specification states the
188.8.131.52 Carrier frequency accuracy. The carrier frequency accuracy at the
antenna feed shall be within 1 KHz of the intended value for all RF
carriers. Recalibration intervals to maintain this accuracy shall not be
less than 90 days.
This is on an Army satellite terminal. Another branch of the Army came up
with the 11 Hz requirement over 24 hours (for testing purposes), which is
just 1KHz/90, and I feel it is total rubbish. I myself have been trying to
find out the gate time used on the counter, but have not been abler to get
it yet and not sure I will. Needles to say, using the stated counter, and
logging a measurement at one minute intervals with a resulting readout down
to 1 Hz at 30 GHz, probably does not leave too many options for gate time.
The jumps do not exactly come in pairs, although they seem to, but, remember
there are close to 16,000 points on that graph. I agree with you that
averaging would make the plot into a straight line.
More information about the time-nuts