Hi Martyn,

I think a few 10s of ps 1sigma error on the 1pps output should be
achievable.  If for example the Rb was free running, and had for example a
100MHz (instead of 10MHz) output, it is not too difficult to measure the
PPS arrival time of the GNSS PPS down a 10ps uncertainty relative to the Rb
reference.  Doing this over a long interval, and fitting the history,one
should be able to predict when relative to the Rb oscillator the next PPS
should occur. This would be in the form of a clock cycle number and
fraction.  Then you should be able to create the PPS output by delaying the
appropriate cycle by the fraction.  One method would be by using a DAC to
control the threshold of something like the ADCMP-580  and creating a ramp
on the other input.  I am 100% certain of the possibility of measuring to
10ps or less, but only theorize the delayed output can be produced,  There
are also delay vernier chips, but I am not familiar off the top of my head.

--mike

On Wed, Dec 4, 2019 at 3:41 PM martyn@ptsyst.com wrote:

Hi

Taking “a few ps” to mean 1 ps (which is not what was said),
you likely need to have a pulse to pulse performance that was
good to << 1x10^-12 at 1 second. First step is to get a couple of
standards that have an ADEV below 1x10^-13 at 1 second. How
much below that level is going to depend on a few things. ( =
how much of the budget is used up by other parts of the system,
how well ADEV relates to what you really need …).

My guess is that the cost of those standards is a non-starter
for most applications.

Bob

Hi,

While being very close geographically, with all the benefits that gives
to common view aspects, I think GPS/GNSS two-way methods will not be
reasonable on their own to achieve precision within handful ps.

Sub-ns takes quite an effort, and the close proximity might help to push
it a little down, but getting it that low I guesstimate as
difficult/impossible. I would say you would need another system for that.

If you can use fiber, have a look at White Rabbit.

Cheers,
Magnus

On 2019-12-04 10:40, martyn@ptsyst.com wrote:

Martyn,

They should look at their best 1PPS on a 'scope. You can get ns with
care; I doubt ps is possible. I mean, that's THz BW isn't it?

Can you share with us what their application is?

Right. It will be ns, not ps. Forget about using GNSS for ps level timing.

Google for papers by high-end national timing laboratories. Words like:
active temperature stabilized (phase stabilized) bidirectional optical
fiber links.

Very possible, very expensive, quite common now. I'd guess most of the
timing centers in Europe are linked this way.

/tvb

On 12/4/2019 1:40 AM, martyn@ptsyst.com wrote:

Hi,

On 2019-12-04 16:55, Tom Van Baak wrote:

No, you do not need THz BW for ps level timing precision, you can
achieve it in the microwave region for sure. Stabilizing it to that will
be a pain regardless, but doable given that the media and environmental
conditions is good enough.

This is indeed a good question. What is the actual requirements and
conditions.

Sub ns is possible but painful, but you are not deep sub ns. This is a
very well studied issue for national timing laboratories and part of the
fundament keeping these labs tied to the major labs to compare for the
full EAL/TAI/UTC time-scale contribution. Major labs also have two-way
satellite links, which is better.

Well, Europe has come further on the optical links than US, but not a
dominant feature, even if the ambition is there. The map shown regularly
looks somewhat more connected than reality gives. I go to these
conferences, and see the progress. Also, there is two types of links,
the frequency links (now stable down to 1E-19), and the time-links. For
time you end up with lots of issues that you can just ignore as you do
frequency, essentially you need to re-learn. For distances like these,
look at White Rabbit for sure. It is being run for some links on long
distances and it is pretty good, but for big-scale usage it does not
make economical sense.

US/North America has extremely little links being done. The only one I
know was really reported was from GPS(MC) to UTC(NIST) local to
Colorado. As I recall it, that was only a temporary setup.

Cheers,
Magnus

Also, at that speed, signal travels 0.3mm per second, assuming zero velocity factor.....  Everything will have to be on chip scale to realize that.

(Mr.) Taka Kamiya
KB4EMF / ex JF2DKG

On Wednesday, December 4, 2019, 11:01:12 AM EST, Tom Van Baak <tvb@leapsecond.com> wrote:  

Martyn,

They should look at their best 1PPS on a 'scope. You can get ns with
care; I doubt ps is possible. I mean, that's THz BW isn't it?

Can you share with us what their application is?

Right. It will be ns, not ps. Forget about using GNSS for ps level timing.

Google for papers by high-end national timing laboratories. Words like:
active temperature stabilized (phase stabilized) bidirectional optical
fiber links.

Very possible, very expensive, quite common now. I'd guess most of the
timing centers in Europe are linked this way.

/tvb

On 12/4/2019 1:40 AM, martyn@ptsyst.com wrote:

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.

Maybe Grace Hopper's aid will help? 11.8 inches = one nanosec, so a ps
is .0118 inches. You really want two clocks not occupying the same space
to be correlated to that accuracy?
Don

On 2019-12-04 08:55, Tom Van Baak wrote:

--
Dr. Don Latham  AJ7LL
PO Box 404, Frenchtown, MT, 59834
VOX: 406-626-4304

On 12/4/19 8:49 AM, Magnus Danielson wrote:

Think of this as a "phase knowledge" question.  The phase uncertainty of
a sine wave in noise is related to the SNR.

In time terms tuncert = 2pif * sqrt(1/SNR)

So you can compute the noise (probably mostly thermal noise in the
measurement band), you know your signal level, and then you're all set.

For, say, 100 MHz reference, SNR = -20*log10(628E6 * 1E-12)  = 64 dB

I totally messed up what I wanted to say.  Here's the corrected version.

In 1 ps, signal travels 0.3mm.  Everything has to be on one die (as in die in a chip) as propagation delay in cabling alone will add enough errors.

(I'm sorry tvb.... for creating more job for you to do)

(Mr.) Taka Kamiya
KB4EMF / ex JF2DKG

On Wednesday, December 4, 2019, 3:07:34 PM EST, Taka Kamiya via time-nuts <time-nuts@lists.febo.com> wrote:  

Also, at that speed, signal travels 0.3mm per second, assuming zero velocity factor.....  Everything will have to be on chip scale to realize that.

(Mr.) Taka Kamiya
KB4EMF / ex JF2DKG

    On Wednesday, December 4, 2019, 11:01:12 AM EST, Tom Van Baak tvb@leapsecond.com wrote: 

Martyn,

They should look at their best 1PPS on a 'scope. You can get ns with
care; I doubt ps is possible. I mean, that's THz BW isn't it?

Can you share with us what their application is?

Right. It will be ns, not ps. Forget about using GNSS for ps level timing.

Google for papers by high-end national timing laboratories. Words like:
active temperature stabilized (phase stabilized) bidirectional optical
fiber links.

Very possible, very expensive, quite common now. I'd guess most of the
timing centers in Europe are linked this way.

/tvb

On 12/4/2019 1:40 AM, martyn@ptsyst.com wrote:

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
 

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.