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

John Ackermann N8UR jra at febo.com
Wed Mar 9 12:40:58 EST 2005


Brian Kirby wrote:
> John, when I was in the Air Force, we had to make fine adjustments to 
> the cesiums, per USNO's orders.  It took several days for the effects to 
> start to show up, and sometimes a week or two, to really see them.
> I would also say you need to increase your averaging time.  If I read 
> you right, you said you can resolve 2 nS in 100 sec, thats only 2x10-11, 
> you need to be a magnitude lower.

Hi Brian --

Thanks, it's really helpful to know about the long time for effects to 
show.  That could explain a lot.  Apparently I'll have to be much more 
patient and let things run for a couple days between tweaks.

Re the averaging, my earlier response may have been confusing as Chuck 
raised a similar point.  I'm doing 1pps comparisons and doing a time 
interval average (using the counter's math) to get a reading every 100 
seconds.  But I'm logging that to a data file and plotting the results 
over much longer times; I'm not basing my results on individual samples.

The main reason for the averaging isn't to get additional resolution 
(though it does effectively increase the resolution from 2ns to 200ps), 
but to average the GPS noise down enough to be able to discern the 
slope.  The raw GPS data would show noise of probably 100ns peak-peak, 
where the averaging brings that down to 10ns or so.  I have a 5370B 
counter that can measure 20ps single shot, but the GPS noise is so 
dominant that the higher resolution doesn't help, so I'm using an HP 
5334A for this exercise.

Here's the current plot that I'm running, automagically updated every 15 
minutes:  http://www.febo.com/time-freq/plots/hp5061a_1-gps.html.

John



More information about the time-nuts mailing list