[time-nuts] synchronization for telescopes

[Attila Kinali]

Moin,

Let's quickly recap what the requirements are and what has been discussed
so far:

1) Time sync system of two (multiple?) sites spaced around 2km to better
   than 1ns, preferably 100ps.

2) System has to be mobile, no fixed installation

3) No amateur radio license available

4) Total cost less than 2000€

5) FPGA skills available


I think the free space laser sync system is a neat idea, but designing and
building it takes quite a bit of skill and experience and it requires
line of sight.

The sync to radio signal would require quite some engineering as well
for anything but AM stations.

Setting up a transmitter doesn't work either because that requires and
amateur radio license.


What I think has the best chances of success is to use an Rb frequency
standard at each site instead. This will give you a stable reference
frequency which will allow you to average the data from the GPS module
to find the precise time in the prostprocessing.

As a GPS module, I would either use an LEA-M8F or a LTE-Lite. The LEA has
an frequency/phase input which which an external reference can be measured.
the LTE-Lite supports using an external oscillator. What you definitely
need is to get the satellite phase data ouf of the module to relate
the phase differences between the modules local oscillator to the satellites
and from there to the other locations.

As for the time stamping. Using a 400MHz clock is ok, but you need higher
resolution for the post processing, as you will need some bits to burn.
As you will be using an FPGA anyways, I recommend using the OHWR TDC Core [1].
This will give you something in the order of 100-200ps resolution.
The CERN people did an implementation on Spartan6, and our group ported it
to Cyclone4 (code not released yet, needs some serious clean-up). From the
problems we faced, i would recommend going for the Xilinx, as the Cyclone4
requires you to get the paid version of Quartus in order to place the circuit
where it works best. Good thing is though, that the FPGA TDC is very stable
over time, the drift we saw in our measurements is orders of magnitude below
the smallest cell size (we measure less than 1ps over 24h... if we can trust
the data)

This should bring you at least down to a 1ns uncertainty level
(after calibration). Judging from Michael Wouters said, probably
close to 200-300ps.


The hardware looks pretty simple: Feed the Rb's output to the GPS module.
Feed the PPS from the GPS and a PPS from the Rb to the FPGA and
time-stamp both (alternatively, use the Rb as frequency reference for the
FPGA). You need be carefull with the power supply of the FPGA and the
signal feeds as now the FPGA is a semi-analog device. Feed all timestamp
data from the FPGA over USB or Ethernet to a PC and store it there.

My guestimate for the cost of this system would be:

200.- for the Rb (from ebay)
200.- for the GPS module
500.- for the rest of the electronics including PCB

The rest is a software problem ;-)


			Attila Kinali

[1] http://www.ohwr.org/projects/tdc-core

-- 
Reading can seriously damage your ignorance.
		-- unknown