[time-nuts] Different Thunderbolt versions

Markus Kern mkern at fastmail.fm
Fri Feb 27 23:13:15 UTC 2009


On 26.02.2009, 20:53 Bruce Griffiths <bruce.griffiths at xtra.co.nz> wrote:

> Markus

> If one used a GPS timing receiver like the M12M + T or equivalent a
> single shot PPS phase error measurement resolution of 1ns or better is
> desirable.
> This is easily achieved using a microprocessor with built in charge
> redistribution ADC and a simple interpolation circuit.

Can you describe what you mean by interpolation circuit? I assume you
are talking about a system which is able to get the 1ns resolution
without requiring the 1 GHz clock necessary with direct sampling?

> One can either devise a means of adjusting the rubidium frequency via
> the C field or use a high resolution, low noise, low spur, synthesiser
> with a small tuning range to produce an output frequency that is locked
> to the GPS receiver PPS output.

> With an appropriate receiver one could even use carrier phase data to
> reduce the measurement noise potentially allowing shorter loop time
> constants to be used.

> Bruce

> Markus Kern wrote:
>> On 24.02.2009, 21:37 Bruce Griffiths
>> <bruce.griffiths at xtra.co.nz> wrote:
>>
>>   
>>> Markus Kern wrote:
>>>     
>>>> On 22.02.2009, 21:12 Bruce Griffiths
>>>> <bruce.griffiths at xtra.co.nz> wrote:
>>>>
>>>>   
>>>>       
>>>>> Markus
>>>>>     
>>>>>         
>>>>   
>>>>       
>>>>> Even with sawtooth correction the performance of the M12+T was found
>>>>> inadequate for the LOFAR
>>>>> <http://www.lofar.org/p/systems.htm> array.
>>>>> They use SRS FS725 rubidium sources disciplined by M12+T GPS timing
>>>>> receivers.
>>>>>     
>>>>>         
>>>> I didn't mean using the M12 by itself, obviously a clock stable enough
>>>> over the time the M12 pps must be integrated has to be used.
>>>>
>>>> If we are using the ADEV limits you proposed then at 50 MHz (= 3ns
>>>> acceptable error) the timing requirement is an ADEV of 3*1E-(8+x) at
>>>> tau = x seconds. From the measurements at
>>>> http://www.leapsecond.com/pages/gpsdo/ it seems the Thunderbolt gets
>>>> pretty close to that.
>>>>
>>>> The LOFAR clock system is described at [1]. In section 3.1.3.3 they
>>>> say:
>>>>
>>>> "Some Crystal Oscillators have the advantage that they have a better
>>>> Allan variance for periods of up to 10s and therefore it can be claimed
>>>> that they have a better performance than the SRS-FS725 Rb-reference
>>>> standard. The performance for time periods above 10s, the SRS-FS725
>>>> performs better. Therefore choosing an OCXO would require a maximum
>>>> calibration interval of 10s and it would require a significantly better
>>>> GPS (or GALILEO) receiver because de Rb-reference is used to average
>>>> the PPS signal from the GPS receiver thereby making it possible to
>>>> identify the time difference between stations at receive frequencies
>>>> above 10MHz."
>>>>
>>>> I think this means that they are using pps integration times above 10
>>>> seconds. I couldn't find any reference to the actual value though.
>>>>
>>>> LOFAR is also working at frequencies up to 240MHz so the timing
>>>> requirements are definitely higher. They say that a station time
>>>> offset of 200ps does not affect performance as long as it remains
>>>> stable over time.
>>>>
>>>>   
>>>>       
>>>>> They also state that the ionosphere contribution to ADEV is about 8E-12
>>>>> @10s.
>>>>>     
>>>>>         
>>>> Yes, from which they infer that "the reference clock shall have an
>>>> Allan variance of 1e-11 or less over 10s." I am not sure if this has
>>>> to do with the propagation of the GPS signal or if they mean that they
>>>> need a clock stable enough to later compensate for the different
>>>> delays of the observed signal through the ionosphere.
>>>>
>>>> Markus
>>>>
>>>> [1]
>>>> http://www.lofar.org/operations/lib/exe/fetch.php?id=public%3Adocuments%3Alofar_documents&cache=cache&media=public:documents:19_detailed_description_of_clock_sync.pdf
>>>>
>>>>
>>>>   
>>>>       
>>
>>   
>>> Markus
>>>     
>>
>>   
>>> The ionosphere contribution to the Allan deviation at GPS frequencies is
>>> much smaller (by a factor of 10-100 or so) than that, as is evident from
>>> carrier phase measurements.
>>> At 50MHz the ionospheric phase shift, dispersion and instability will be
>>> much greater than at GPS frequencies.
>>> They are merely ensuring that the LO contribution to Allan deviation is
>>> much smaller than that of the ionosphere.
>>>     
>>
>> Yes, that's what I thought.
>>
>>   
>>> If you look at the Allan deviation plot on the PRS10 page:
>>> http://www.thinksrs.com/products/PRS10.htm
>>>     
>>
>>   
>>> This indicates that the likely disciplining loop time constant will be
>>> several thousand seconds.
>>>     
>>
>> Ok, so it may indeed be necessary to use a rubidium oscillator which
>> has the required stability over that time frame.
>>
>> I suppose a GPS disciplined Rb-clock will be much more expensive than
>> a Thunderbolt. However there are relatively cheap rubidium oscillators
>> like the LPRO 101 out there. Are they suitable and has anyone tried to
>> slave them to GPS? From reading the mailing list archives it seems
>> Brooks Shera's circuit won't be suitable for this.
>>
>>   
>>> Close isn't good enough: the phase differences between pairs of stations
>>> is significant, the Allan deviation needs to be at least 30% lower per
>>> station.
>>> If the errors at  station pairs have significant correlation the
>>> requirement can be relaxed somewhat.
>>>     
>>
>> I realize that the phase difference is important and that things won't
>> work if it's not low enough :)
>>
>>   
>>> Bruce
>>>     





More information about the time-nuts mailing list