New Paper

Evaluation of long term performance of continuously running atomic fountains. Metrologia 51 263-269

An ensemble of rubidium atomic fountain clocks has been put into operation at the US Naval Observatory (USNO). These fountains are used as continuous clocks in the manner of commercial caesium beams and hydrogen masers for the purpose of improved timing applications. Four fountains have been in operation for more than two years and are included in the ensemble used to generate the USNO master clock. Individual fountain performance is characterized by a white-frequency noise level below 2 × 10^−13 and fractional-frequency stability routinely reaching the low 10^−16 s. The highest performing pair of fountains exhibits stability consistent with each fountain integrating as white frequency noise, with Allan deviation surpassing 6 × 10^−17 at 10^7 s, and with no relative drift between the fountains at the level of 7.5 × 10^−19/day. As an ensemble, the fountains generate a timescale with white-frequency noise level of 1 × 10^−13 and long-term frequency stability consistent with zero drift relative to the world’s primary standards at 1 × 10^−18/day. The rubidium fountains are reported to the BIPM as continuously running clocks, as opposed to secondary standards, the only cold-atom clocks so reported. Here we further characterize the performance of the individual fountains and the ensemble during the first two years in an operational environment, presenting the first look at long-term continuous behavior of fountain clocks.

I bolded something I harp on occasionally: these clocks are actually run as, and are reported as, clocks. For all of the awesome performance of other devices that grab pop-sci article space, they don’t run continuously and aren’t described as clocks when it comes to the data that get reported to the international standards lab.

Ask Ethan #30: Long-term timekeeping

In this week’s Ask Ethan, we take on perhaps the longest question of them all, and look at how to keep time for arbitrarily long times.

It’s a good post as usual, though there are a few things Ethan glosses over, which is where I step in.

[millisecond pulsars] are also the most accurate clocks we’ve ever discovered. They are so regular that we could watch one, look away for a year, and know — when we look back — whether ten billion pulses have gone by… or whether it’s ten billion-and-one. In fact, we can get down to around microsecond accuracy to their timing over periods of many decades, meaning we can get timing accuracy to around one part in 10^15!

This is bettered only by the most advanced atomic clocks on Earth

In terms of fractional stability that’s true (and I think he means precision rather than accuracy here), but Tom- man-made atomic clocks reach this stability in a matter of hours or days, not years. It’s only by having these good clocks that we can measure how well the pulsars are doing.

I have a recollection of a discussion about timing with pulsars from years ago (this isn’t a new idea). Pulsars don’t actually have an inherently stable frequency — they are slowing down, just like other macroscopic spinning objects, so the timing will show a drift. But pulsars do this at a very predictable rate, so you can characterize them and account for the drift. Some pulsars haven’t “settled in” and can undergo a star quake, which changes their rotation abruptly, but I think the ones under discussion are past that age.

From his followup post

Atomic clocks require a lot of power to continuously stimulate atomic transitions, and a lot of cryogenic fuel to keep the atoms at ultra-low temperatures. Not such a big deal when you’re talking about doing this in a continuously powered laboratory on Earth, but that’s a lot of resources to devote to keeping a simple clock running. The mechanism I gave in the original article — counting atoms or looking at a pulsar — has the advantage that all you have to do is look once at the beginning and once at the end, and requires no devoted power in the intermediate time.

The cryogenics part is a common misconception, since the best clocks are cold-atom clocks. So the assumption is understandable, but these clocks are all offshoots of laser cooling and trapping. Some crystal oscillators in frequency standards are cryogenic, but not in any continuously-runnung clock. That’s a logistical nightmare.

I think our clocks are not getting quite as much respect as they deserve — the pulsars have to be characterized to be useful, and no two pulsars are going to have the same frequency. You could, in principle make a stable reference from measuring several of them, but to actually tell time you have to tie it back to a standard, which means a man-made device.

The second part, about radioactive decays, stresses the longevity of the clock but ignores the precision of the measurement. Unless you have a huge chunk of the material and can determine the number of atoms precisely, your counting statistics will limit the precision of your measurement.

The upshot of all this is that timing is a little more subtle than having something that will tick for a long time. Accuracy, precision and durability aren’t interchangeable attributes. There are often tradeoffs between each of them, so it depends on which is more important.

A couple of recent articles on timekeeping, which are always fun.

First, medieval timekeeping
Tempus Fugit

and second, the birth of atomic time
Essen’s clock

New atomic clock could keep accurate time until the world ends

\(\Psi\) (Sigh)

If the atomic clock in the University of Colorado Boulder’s JILA laboratory had been started when the earth came into existence, its time would still be perfect down to the very second today. Likewise, if the clock were reset now and kept running, it would likely outlast life on Earth.

Aye, there’s the rub. …and kept running. But it doesn’t — it only runs for a few hours. Which means my standard disclaimer applies: this isn’t a clock, it’s a (kick-ass awesome) stopwatch.

So when Jun Ye says “You can expect more new breakthroughs in our clocks in the next five to ten years.” what he means is that they will continue to push for even more stable clocks — greater levels of precision and, if these are going to become primary or secondary frequency standards, greater accuracy. They are not going to to be pushing the envelope with respect to robustness of the technology unless it furthers the goal of better accuracy/precision — that’s not really their job. It’s my job. (Yeah, I’m starting work on an optical clock)

A confusing this is that they mention a competitor’s clock and claim it uses Cesium, but the linked article says it also uses Strontium. When it says it measures time in a “non-standard and still unaccepted way” I think they are referring to the fact that it’s an optical transition, and not at the ~9.192 GHz transition that defines the second. But non-standard and unaccepted? Not so much — the Rubidium fountains I have helped build, and Hydrogen masers that are in widespread use don’t/can’t rely on that transition, and these clocks are reported to the international Bureau of Weights and Measures.

Those stumbles aside, the applications mentioned at the end — precise sensors for gravity, for example (I think “quantity” in that bit is an autocorrect casualty and is supposed to be “quantum”), but not so much for timekeeping.

Title reference

No, this is not one of the pretenders that link up to NIST’s atomic time via a radio signal.

$12,000 watch has its own built-in atomic clock

I link to this article because it actually mentions USNO, but there’s the original, which mentions it’s made (or will be made) in Switzerland, meaning this is probably not just a Symmetricom CSAC that’s been marked up with a counter and a display attached, but it’s undoubtedly the same technology.

This watch actually points to a problem in timekeeping, that there are two elements one must worry about: telling the time, and disseminating the time. Having a great clock is not particularly useful if you can’t transfer the information to anyone, so there is a dual, usually parallel effort to improve clocks and to improve time transfer. Time transfer can’t lag too far behind timekeeping or else there’s no point in pushing the boundaries.

Here we have the time transfer problem in reverse. If the input is the stem and you have to look at a display (or listen to a voice) to get the time, it is going to be limited to the feature of not gaining or losing a whole second over some long interval. Which goes out the window because you have to reset it when you change the batteries. The watch really doesn’t require or exploit its precision, so why? It’s really nothing more than an expensive trophy, while some pretty incredible technology is basically wasted. And an analog display? I’d want a digital one that showed the time to better than a second.

However, this does point out the ridiculousness of an episode of Person of Interest from last season, where a very rich guy™ supposedly had a watch that kept time to the nanosecond. 1 second in 1000 years is roughly a part in 10^10, so that’s not even a microsecond per day.