[time-nuts] Cold Rubidium over hyped?

[Attila Kinali]

Moin,


On Fri, 8 Nov 2019 14:33:45 -0800
<cdelect at juno.com> wrote:

> Well I have been looking at the data in the 
> 
> https://www.researchgate.net/publication/325499937_A_portable_cold_87_Rb_
> atomic_clock_with_frequency_instability_at_one_day_in_the_10-15_range
> 
> link and find that maybe they are overstating their performance!
> 
> In the plot of figure 5 my HP 5065A almost perfectly matches the Allan
> deviation 
> out to 1000 Sec.! (Although mine does have VERY good performance)

And you don't think this is impressive? Short term performance of an
atomic clock depends on the SNR of the signal. Which in turn depends
on the number of atoms being probed. A vapor cell standard probes
billions of atoms even for tiny cells, 1e15 to 1e18 for large cells.
A MOT usually traps a few 10k to at most a few  tens of million of atoms.
Ie we are at least a factor of 1000 below a Rb vapor cell, and more like
a factor of billions. And as usual, noise goes down with the square root.
Which in turn means the theoretical SNR in a cold atom Rb standard is a
factor of 10'000 lower than your 5065! Yet it achieves the same short
term performance! (The real differnence in SNR is probably lower than
that, but I doubt it's less than a factor of 100)

And mind you, we are still talking about quite new products where little
optimization has been done. And we aren't talking about big companies
like HP either. µQuans was started by a few students, who made their 
research into a product. SDI is a very small business. Neither
have millions to spend on optimizing the product until it's perfect
before they announce it to the world. They had to get a product out
as soon as possible to make sure they can recapture the cost of development.
And don't worry, both are working on improving the system.

To make things worse, the paper you are citing is actually the first
one SDI published. Meaning that was (one of?) the first complete system
they had built.

 
> Also if you look at figure 7 all the Maser data shown seems to be from
> poorly
> operating Masers so any judgement of better clock performance versus the
> Masers
> is a bit much!

Calling the masers of NIST poorly maintained is a bit much, isn't it?

Besides, they do not measure against the masers, but against UTC(NIST),
though that's not that clear from the paper, as they only say "NIST
measurement system." Other papers state it explicitly.


> Something does not add up in their data!

I think it clear from what I wrote above and what others wrote before me
that you are comparing apples and oranges. While it is true that both
the cold atom clocks and your 5065 use Rubidium as a species, the techniques
involved are so differnt that you have to treat them as a totally different
kind of clock. If you look at them as a active H maser or Cs beam standard
replacement, that would be closer to what they are.


On Sat, 09 Nov 2019 00:07:43 +0000
"Poul-Henning Kamp" <phk at phk.freebsd.dk> wrote:

> Given that these two cold-Rb devices are 1st generation of their
> kind and given how little time they have had to collect data on
> them yet, both with respect to performance but also day-to-day
> gremlin-wrangling, I wouldnt be at all surprised if the next
> two generations of that concept delivers almost two orders of
> magnitude improved performance.

While both companies are working on improvements. I don't expect
two orders of magnitude. Both systems work close to what the
theoretical limit of the number of atoms they are probing are.
I wouldn't be surprised if they get a factor of 10 out of the
system, after all improvements have been done. But I would find
it unlikely that the reach a factor of 100 without significantly
increasing the number of atoms they probe or increasing the
interogation time. And both of these things are quite hard to do
in these architectures.

On Sat, 09 Nov 2019 07:05:10 +0000
"Poul-Henning Kamp" <phk at phk.freebsd.dk> wrote:

> It is not like people have not been improving the Rb-Gas-cell, but if
> you want to improve it an order of magnitude all the myriad of "small
> issues" to fix add up real fast, and when you are done with all of
> it, you will still have frequency drift.


If you want to know what the limit of Rb vapor cell standards are,
have a look at the research done by the group of Mileti and Affolderbach.
They are at ~1e-13/sqrt(tau) and are trying to push it down to the shot
noise limit (which is at a few parts in 1e-14). That's the limit you
can do. And that's still not accounting for drift and other things which
limit the ADEV to a few parts in 1e-14 somewhere around a tau of 1k to 10k.



On Sat, 9 Nov 2019 02:38:45 +0100
Magnus Danielson <magnus at rubidium.se> wrote:

> This is well understood techniques now, so that it matures into
> commercial products is not strange. It's worth noting that optically
> probed cesium also exists from Oscilloquartz, altering the beam standard
> techniques

It is also worth noting that it took Oscilloquartz several years to get
it working. And not only that, a lot of people in industry believed that
it would not be possible to make a fully optical Cs beam standard that
is as reliable as the 5071 and at the same performance. As many have tried
and have failed.


			Attila Kinali

-- 
<JaberWorky>	The bad part of Zurich is where the degenerates
                throw DARK chocolate at you.