[time-nuts] Checking accuracy of Rubidium standards
Neville Michie
namichie at gmail.com
Mon Nov 10 02:55:36 UTC 2008
Bruce, how does the three cornered hat scheme work?
If I had two LPRO Rubidium oscillators and a TBOLT GPSDO, and I
divided each of them down to
100KHz, then I could compare pairs of them with D latches and record
3 different analogue
signals of phase difference.
One of the LPRO oscillators is in an oven to remove ambient
temperature influence.
If I ran them for several weeks and logged the signals every 10 minutes,
what could I expect to recover from the data and how would I apply
the 3 cornered hat scheme?
I ask this question because this is about where my building program
is taking me.
cheers, Neville Michie
On 10/11/2008, at 1:38 PM, Bruce Griffiths wrote:
> Joe
> J. L. Trantham wrote:
>> I have been enjoying this discussion.
>>
>> Since the original question was the desire to 'compare' the
>> frequency of an
>> LPRO to a Z3801, it seems that you could consider that from two
>> (at least)
>> perspectives.
>>
>> Before I begin, I confess that I am a novice in this arena and please
>> correct me in any area that needs it.
>>
>> The first perspective is the issue of frequency. That seems to me
>> to be the
>> issue of the average frequency of the LPRO versus the average
>> frequency of
>> the Z3801. Assuming that there is no gross difference of the 10 MHz
>> signals, a lissajous figure (X-Y display) on a scope with the
>> appropriate
>> bandwidth amplifiers would be a reasonable initial approach.
>>
>>
> The limitation with using Lissajous figures is that eventually the
> noise
> in the relatively wide oscilloscope bandwidth (due to amplifier noise
> and signal phase noise) limits the figure rotation rate that can
> reliably discerned. Also the lack of a record of changes limits the
> ability to see long term trends/drifts.
>
> However it can be a useful/instructive starting point particularly
> when
> the frequency difference is relatively large but still small enough
> that
> the figure rotation can be seen.
>> Assuming that they are both near 10 MHz and you do not know which
>> is the
>> most accurate (although the Z3801 would seem to be the default
>> standard), if
>> it takes 10 minutes for a single cycle of the lissajous figure to
>> complete,
>> then it is 1 cycle per 600 seconds difference between the two and
>> therefore
>> the two are within 1/600 Hz or 1.67 mHz of each other. If we
>> assume that
>> they are both close to 10 MHz, then that is 1.67 parts in 10E-10
>> difference
>> between the two. Is my logic faulty?
>>
>> The other perspective is the issue of 'purity'. That is to say,
>> what is the
>> 'frequency modulation' of the source? This, I think, is the issue
>> of phase
>> noise. Correct?
>>
>> That is something that I have not yet had a chance to contemplate
>> as far as
>> how to measure. It would appear to require a particularly stable
>> (pure)
>> source as a reference though. Various multiplying or dividing
>> protocols
>> would seem to introduce a host of other variables that would seem
>> to be
>> difficult to account for though they might accentuate an impurity
>> in the
>> signal in question. I have read Bruce's comments and I still do not
>> understand the basics of time stamping or how a sound card might
>> provide
>> this.
>>
>>
> One technique is to use a three cornered hat technique.
> If the phase fluctuations of 3 oscillators are statistically
> independent, then it is possible to determine the statistical
> fluctuations in each of them.
>> I would appreciate any direction for further reading regarding
>> this and I
>> would appreciate any direction/correction/etc. in the thoughts above.
>>
>> Joe
>>
>>
> Bruce
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/
> time-nuts
> and follow the instructions there.
More information about the Time-nuts_lists.febo.com
mailing list