[time-nuts] Different Thunderbolt versions

[Markus Kern]

On 22.02.2009, 18:33 Hal Murray <hmurray at megapathdsl.net> wrote:

>> I actually want to work at frequencies below 50Mhz where a 50 degree
>> phase error corresponds to about 3ns.

> What is interesting to listen to below 50 MHz?  Are there any handy strong
> signals you can use for debugging/calibration?

There are several bright sources which should be relatively easy to
observe: Cygnus A, Virgo A, Centaurus A, Taurus A, Fornax A and of
course the sun.

>> Also at these low frequencies the required accuracy approaches that
>> claimed by some GPS receivers (4ns for the M12+ ?) so might it be
>> possible to build a system which is accurate for arbitrary integration
>> times? 

> If you haven't seen it yet, this is wonderful background:
>   http://www.gpstime.com/files/PTTI/PTTI_2006.pdf
> In particular, be sure to check out the hanging bridges.

> How far apart are your antennas?  Things might get a lot simpler if you can
> see the same GPS satellites.

I haven't decided on a fixed baseline. I plan to use a semi-mobile
station to increase the baseline once it works at closer distances. So
using the same satellite might be possible but I guess there are
additional complications of synchronizing the handover in that case.

> If your runs are short enough, you might get better results by letting the
> oscillator coast during the run.  That is unplug the antennas or turn the
> feedback off in software.

Yes, that might be something worth trying.

> This could be all wrong....

> There are two things you need.   One is frequency, the other is time (PPS).

> If the time is slightly off, that's equivalent to pointing your antenna in a
> different direction.  You get the same sort of error if the location of an
> antenna is slightly off.

> I think the Earth's rotation introduces an error that's equivalent to the
> frequency being slightly off.

I think there are two requirements. One is having the sampling
synchronized between stations (I want to use direct sampling without
any analog mixing) to be able to correlate the signals at all. The
other is tagging the samples with the current time in order to keep
the search space for the correlator sufficiently low. I believe the
later can be easily achieved by synchronously sampling the pps signal
together with the data.

Markus