[time-nuts] Optical transfer of time and frequency

[Magnus Danielson]

Hi,

On 04/29/2016 11:45 PM, Michael Wouters wrote:
> On Sat, Apr 30, 2016 at 6:14 AM, Magnus Danielson
> <magnus at rubidium.dyndns.org> wrote:
>> Well, giving the conditions mentioned, doing ranging codes such as those
>> used by GPS is very easy and cheap. Doing this in bidirectional isn't too
>> hard. Doing a suitably high chip-rate should cost very little.
>
> I've done two-way time-transfer over optical fibre using exactly this
> technique. The TDEV is about 1 ps for tau>1s. Not so cheap, about 25K
> euro per node (20K signal processing - NI FPGA, 2K laser and power
> supplies, 1K detector, 1K RF electronics) in my setup, but that cost
> could be greatly reduced since a $100 OEM FPGA could do the signal
> processing (I've already done work on this but currently looking for
> motivation to finish it off) and a simple, intensity-modulated laser
> would probably be fine. A 2K euro budget would be a challenge though.

FPGA-wise, you need a very little FPGA resources.
If you consider the RedPitaya (200 USD) for instance, it is way beyond 
what is needed.

>> The two-way time-transfer is relatively easy, but you will need to do some
>> calibration to get the precision needed.
>>
>
> At first glance, I would think that you should be able to define the
> optical RX/TX path to within 10 cm without any trouble and that gives
> you 300 ps accuracy. Even on fibre links, I don't think anyone would
> claim an accuracy of better than a few hundred ps.

With a bit of calibration you can remove each nodes systematic asymmetry.

For optical fibers many does not even bother to do a pseudorandom 
rangning. A repeating pattern suffice, such as that of SDH frames.

Cheers,
Magnus