[time-nuts] How close can you trim a Cs?

[Brooke Clarke]

Hi John:

I too have been going nuts (maybe there's more than one meaning to the 
name of this list server) what you describe.  I think I have a handle on it.

Assume that you had the C field tweaked perfectly and then ran the test 
you are now.  I would expect to see random walk excursions away from and 
back to perfect.  Now assume that the oscillator is close to perfect.  
Now when the random walk variations occur at some point the time 
interval will be perfectly flat for many many hours before it goes back 
to the slope indicting the true offset.

Another way to think of this is that if the random variations were a 
sine wave and they were superimposed on a line with some slope, then 
there will be two points in each cycle where the combined slope is flat.

I find that it takes at least 3 days to see what's going on, a week is 
better.  I just posted the idea of watching the standard deviation on 
the averaged time interval.  This has been very helpful for me both in 
confirming that my 50 degree elevation angle GPS mask is working and in 
catching the FTS4060 loop oscillation after adjusting the beam current 
gain.  Easy to fix the oscillation by cycling power (off for 15 seconds) 
and quick to relock since the oven was warm.  Every now and then I get 
an SDEV greater than 40 ns and just toss the data.

I think as you try to set to lower and lower stabilizes it's going to 
take longer and longer to know if you're there. 

The manual for the FTS4060 says the thumb wheels are 2E-14 per click, 
but I've never seen it work that way.  Like you mentioned it's more like 
change the C field and roll the dice for the amount of change.  Although 
the direction of the change is correct.  That's why I have gone to a 
binary search mode of choosing the thumb wheel setting.   I'm now down 
to within a few clicks of the final value.

In answer to the subject line I think there is some limit imposed by not 
just by the physics but also the design of the standard itself.  So no 
matter how hard you try there will always be some error and it will 
accumulate over time.  But if you characterize it then you will be 
working with a known quantity.

Have Fun,

Brooke

-- 
w/Java http://www.PRC68.com
w/o Java http://www.pacificsites.com/~brooke/PRC68COM.shtml
http://www.precisionclock.com




John Ackermann N8UR wrote:

> I've been going nuts trying to trim my 5061A versus GPS.  I have it 
> well within 1x10e-12, but trying to get down into the low 13s is 
> proving challenging.  Making C fields adjustments of a couple of minor 
> divisions that according to the book should be result in a shift of a 
> few parts in 10e-13 don't seem to have the effect they ought to; in 
> some cases the frequency seems to move in the opposite direction.
>
> For example, I did a run that showed about +3x10e-13 and tweaked four 
> minor divisions (nominally 3.2x10e-13 according to the book), which 
> should have brought it down to nearly flat.  Instead, the offset 
> increased to about +6x10e-13.  So, I adjusted the eight minor 
> divisinos back (so I ended up the at same shift, but opposite 
> direction from the starting point).  Now, I'm reading +7x10e-13!
>
> It may be that I'm being too impatient and not letting the comparison 
> (against raw GPS 1pps) run long enough, but when after 15 hours I'm 
> seeing a plot that shows a pretty stable slope, it's hard not to 
> assume that things need more tweaking!
>
> So maybe the first question is -- at these offsets, how long should I 
> track against GPS before making a tweak?  And the second question is, 
> at what point am I tweaking in the noise -- how close is it realistic 
> to aim for with this class of standard?
>
> Thanks,
>
> John
>
> _______________________________________________
> time-nuts mailing list
> time-nuts at febo.com
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>
>