[time-nuts] suitable statistics for measurements with gaps

Magnus Danielson magnus at rubidium.dyndns.org
Sat Apr 23 19:02:20 EDT 2016



On 04/23/2016 12:10 AM, jimlux wrote:
> On 4/22/16 12:41 PM, Hal Murray wrote:
>>
>> jimlux at earthlink.net said:
>>> But what about when the observations have gaps? Say you're measuring the
>>> frequency of a spacecraft oscillator, and you can only see it for 8
>>> hours a
>>> day?
>>
>> One interesting question...  Can you match up the cycles after the
>> gap?  Is
>> your clock stable enough or do you have a slower clock (PPS?) that you
>> can
>> lock on to.
>>
>>
>
> probably can't match the cycles..
>
> It's a sort of generalized question, but say your clock is a OCXO with
> 1ppb sort of accuracy, that's 1E-9 in a long term sense, and after
> 20,000 seconds of a 10 MHz clock, you'll have accumulated 200E9 cycles.
> If your frequency changed the 1ppb, that's 200 cycles error.

Observing the frequency drift, matching it, making it match might be 
possible. Probably easier on one of those spacecrafts than on my lab 
bench. The oscillators you use won't be complete crap.

> And I think that gets to my underlying question, is such a measurement
> meaningful?  What statistic, other than, say, "frequency error measured
> at 24 hour intervals, with 1000 seconds of counting for each
> measurement", which is sort of like ADEV with tau of 86400, but not really.

Actually, you can make such a measurement for a range of tau which will 
be the ADEV without problems, meaning the same thing.

> And would that really help with understanding of hte underlying
> mechanisms.  For oscillators, knowing that there's a difference between
> 1/f^3 and 1/f and 1/1 phase noise and then seeing the output of an
> oscillator helps you to optimize the overall system design. if the white
> noise floor is really high, then improving flicker noise may not help.

It would be a bit different, but not worse than it can be used.

Cheers,
Magnus


More information about the time-nuts mailing list