[time-nuts] Current state of optical clocks and the definition of the second

Magnus Danielson magnus at rubidium.dyndns.org
Wed Jan 14 00:05:23 EST 2015

On 01/13/2015 11:41 AM, Attila Kinali wrote:
> On Mon, 12 Jan 2015 20:09:45 +0000
> Gregory Maxwell <gmaxwell at gmail.com> wrote:
>> On Mon, Jan 12, 2015 at 12:34 PM, Attila Kinali <attila at kinali.ch> wrote:
>> Seems that the state of the art in stabilized lasers has improved a
>> lot lately, e.g. there are commercial available 1550nm devices which
>> have a <=3Hz line-width: http://stablelasers.com/products.html (well
>> on a short term basis, the medium term performance is not so
>> impressive)
> Laser stabilization, especially for quantum metrology is still
> an actively researched field. Current state of the art is IIRC
> 0.3Hz linewidth (sorry, cannot find the reference at the moment).
> Mid- and long term stability depends highly on the reference
> used. Current research is fucused mainly on special, low vibration
> structures made out of low expansion glass or silicon. And these
> cavities are usually put into a temperature controlled chamber in
> vacuum.

Well, guess what I found standing around in a lab with an optical comb? :)

With optical line-widths in sub-Hz range and optical combs you have a 
nice way of comparing the frequency of that free-running and 
un-steerable but stable oscillator. However, as you mix it down the 
noise of the optical comb will dominate, but you can know which multiple 
of the optical comb and offset it is.

>> Considering the rarity and extreme cost of H-masers, or just really
>> exceptional quarts oscillators; might it be the case that optical LOs
>> start looking interesting for applications which just need stability
>> (or being steered by other sources; e.g. GPSDL)?
> Well, an 8607 costs more than a Rb-standard. Yes, the 8607 has lower
> close in phase noise and up to several 1000s it rivals the Rb, but
> handling it is much more difficult than handling an Rb.
> Also, if you want to buy one of those exceptionally low noise/high stable
> 8607's (those that go down into the 10^-14 range) you'd have to sell your car.
> But, if you buy a H-maser from SpectraTime, you get a 8607 for free ;-)

That is also the only way to get the 8607 now, as Oscilloquartz is 
closing down that business.

> There used to be quite some literature on how to build low noise
> quartz oscillators. Most of those books are out of print today.
> With two notable exceptions:
> "Discrete Oscillator Design: Linear, Nonlinear, Transient, and Noise Domains"
> by Randall Rhea, 2010
> and
> "Understanding Quartz Crystals and Oscillators", by Ramon Cerda, 2014
> I had a look at the book by Rhea, it looks quite well written and contains
> a lot of real world information, but is a bit weak on the more theoretical
> part (description of oscillation, noise sources,...) and thus on the
> on the why things are done that way.
> I didn't had the chance to buy Cerdas book yet.

An interesting book in that context is Enrico Rubiolas book on 
phase-noise, which among other things goes into explain the Leeson model 
of oscillators and it's real life design aspects.

> The UFFC has some of the older books online. You need to be registered
> to access them, though.
> There is also a lot of knowledge on quartz crystalls hidden in old papers,
> but going trough them is some serious work.
> On the topic of opto-electronic oscillators, those are technologically
> nice, but they are rather bulky. That's why they are mostly used in
> research projects for atomic clocks. Also getting them to do low phase
> noise is not that easy, and unlike quartz oscillators, there is not
> much literature about that.

It's a serious bulk of glas in there, but the laser-technology as well 
as temperature-stabilization of it isn't rocket science,

>> (They can be down-converted to microwave frequencies using an optical
>> comb; a mode-locked laser whos pulses are phase locked to an incoming
>> beam.)
> That is actually the current trend. There was a paper by NIST last year
> on downconverting the beat frequency of an optical comb down to RF using
> a frequency divider chain. They managed to get noise measures that rival
> that of a good quartz oscillator at 5MHz. Ie at higher frequencies, it is
> actually better than what a quartz oscillator can deliver.
> (for some reason i have not archived that paper and google fails me)

The NIST T&F archive is where you should go. Nice folks doing that work.

>> Certainly just the local oscillator is _closer_ to something a
>> time-nut might experiment with than a complete optical atomic standard
>> (if still not quite in reach).
> Well, building a CPT based Rb vapor cell frequency standard should be feasible.
> Yes, it's not a primary standard, but should do the job for most :-)
>  From what i've read, using one of the MOT cells like those of
> Sachser Laser [1] one might even be able to build a primary standard.
> But my understanding of MOT is relatively weak and i cannot say how
> difficult it actually would be. But it would be definitly a fun project
> to try :-)

Building a MOT setup is relatively easy know, I have not yet seen it go 
sub 100.000 USD, but it seems like modern setups could do that if you 
let it. I've been playing with such a system and when setup, tuning it 
up and locking it to do cooling is so trivial that I've done it.

> But i agree, building one of those ion or neutral atomic standards is
> pretty much out of question on a hobby budget. Heck, even an optical
> frequency comb is difficult to build, at best. And buying them.. i think
> buying a good Rb is still cheaper.

The technology of such traps is pretty established knowledge, but an 
engineering feat to be done is to have those drop in price, and doing so 
could have very interesting implications.

Once you have your atoms in some form of (glas) cell, is more about 
reducing cost of the optical bench needed. Would be interesting to see 
what some intense "research" could be done in that regard.

Laser cooling depends on the lockable frequency of the lasers as well as 
the detector cavities to make it easy to detect where in the range the 
laser is, once locked to that cavity, a relatively small optical bench 
is needed and well, the magnetic field is just two coils and a 
current-source supply, which ends up not being that critical in setting.
The optical effect for these things seems to be relative benign too. 
Just haven't had the time to learn what (cheap) lasers is available to 
steer, and how the steering is actually done.

Rubidium actually laser-cools very well, better than cesium, because of 
the lower cross-section area of rubidium than cesium atoms, which 
reduces the collision effects of the atoms in the super-cooled ball.
This is just the same rubidium that we have in our rubidium clocks. The 
non-primary aspect of gas-cells is due to wall-shift and gas-shift of 
the buffert-gas to compensate for wall-shift. The temperature of the 
cloud also contribute, but that is way down in the list.

Building a modernized laser-cooled rubidium cell atomic clock would be 
doable. It might not be the cool research project maybe, but an 
interesting step-stone mix of technologies.


More information about the time-nuts mailing list