The coolest part it that it improves both stability and accuracy. Cesium is often touted as a good clock but it only has good accuracy. The short-term stability has more noise than something like a Rubidium clock that is very stable on the short-term but inaccurate(relative to cesium anyway) on the long term.
This is cool because it is the best in both dimensions.
The phase noise (really short term stability) of a rubidium oscillator is rather poor compared to on ovenized quartz oscillator. So what I really want is a strontium disciplined OCXO.
Years ago I was involved in some work on clocks that needed very high performance at low cost. We ended up with OCXOs GPS-disciplined over some of their frequency and atomic clock disciplined (cesium, IIRC) over another part of the range.
Cool. I bought a couple of HP Z3801A recently for home lab use. Also have an Arbiter GPS disciplined clock at work as a frequency standard for my RF equipment; they use one of the Wenzel OCXOs that has the lowest phase noise you can buy.
Some Rubidium clocks require very little power (700mW or less). This makes them ideal for under water applications where short term accuracy isn't an issue because the SONAR data transmission is going to be a mess anyway, and they provide much better long term stability than a crystal oscillator.
That depends on what you mean by accurate, and even then they pretty much suck. Clocks are "inaccurate" in a few ways. One of these ways is very short-term: variations in exactly when the clock edges arrive at frequencies near the clock frequency. Others of these are long-term: how far the clock is from it's target rate over decades. Between these there's a whole spectrum of different noise frequencies.
The oscillators on your motherboard, with very basic temperature correction, very basic calibration and only very basic quality control don't tend to do very well on any of these measures. They tend to do better on the high-frequency measures of clock badness, and worse on low-frequency measures, but they aren't stellar in any area. The end-to-end time system, including NTP and software-level drift correction, does an OK job in some circumstances. It does a horrible job in others.
If you go through life assuming your computer's local clock isn't literally out to get you, you're going to be disappointed.
Quantum leap refers to a sudden increase in something. Given that this clock is 50% more precise than the previous record holder, I think it qualifies to be called a 'quantum leap'. So even if you are pedantic about headlines, it should still make sense. Don't you think so?
It's a Latin noun, derived from an adjective, that literally means "a something of some size." That's it. The first meaning in English is "some quantity." Uses which denote or connote 'large' or 'small' are secondary extensions which depend on context.
"Quantum leap" connotes an unpredictability in the magnitude of the change in common usage. Over time, the understanding has drifted a bit to include, usually, the idea of significant magnitude.
You (and everyone else who is commented in reply to you so far) didn't read the article.
In this particular case it is a perfect use of the phrase, as the advance in accuracy is based on measurements of the atomic electron transitions, also known as quantum leaps.
<random internet commentator tripe>This could be the basis of a future tricorder or Star Trek-like sensor array. Three ultra-sensitive clocks in an array should be able to infer mass and motion both for the unit and objects in the local area indirectly. Relativistic effects are minuscule, but not non-existent. Extremely-tuned clocks would have some pretty cool capabilities.</random internet commentator tripe>
One thing I've spent some time thinking about is whether it would be possible to use atomic clocks to get a better measurement of the gravitational constant G. (G is by far the most poorly known of all the fundamental physical constants.) The idea would be to build a sphere whose mass is very precisely known and place one clock near the mass and one clock far away. By measuring the gravitational time dilation you could infer G. As I recall from my order of magnitude calculations, atomic clocks would have to improve in accuracy by four orders of magnitude or so before this would be feasible. So it would still be a long ways off.
Good point! The research group is still working on refining their system. So may be they will have another breakthrough in a year or two, you never know!
Pulsars are actually pretty noisy. Starquakes, infalling matter and braking due to drag on the magnetic field jink the rotational frequency around a lot, comparable to a good clock (you could probably compensate for the braking effects, but the quakes are another matter).
Build several identical clocks and watch the ensemble drift relative to each other.
The lattice clock in this story has numerous clusters of atoms. They could fill only a few clusters and measure the performance relative to a good clock, then statistically derive how much it would improve by using the full set of clusters.
At this level of accuracy things are pretty strange. It used to bug me when you had to tune RF circuits at a distance with non-conductive tools because your body capacitance would throw the tuning off. Having a clock that just being near it will change what time it reads, well that is a whole different ballgame of weird is it not?
On a science note, why isn't this a gravity wave detector anyway?
Gravity wave detectors get phase accuracy by comparing a photon to itself using an interferometer. SNR is improved by brightening the laser and averaging over more photons. The laser frequency is less important.
It turns out there are some ultraviolet nuclear transitions. The line widths promise to be obscenely narrow. If they can get it working in a clock, they will be able to directly measure gravitational time dilation of small masses.
Do you have a reference to the UV transitions? I'm sure the LIGO folks think of themselves as being in a race with advancing clock technology with regards to actually detecting gravity waves.
I could not find many references, but I think clocks mainly make the instrument cheaper, or possible in the case of long-baseline satellite instruments.
Meanwhile on servers running Windows, keeping sync to within a few seconds is a mighty challenge, esp. with Hyper-V. (Linux guests have no issue with ms or sub-ms accuracy.)
Microsoft's Win32 time is only meant to prevent Kerberos error, so under 5 minutes is "fine".
Do you know of anything more accurate, especially under Hyper-V? I've tried installing ntpd and even using the most aggressive settings and host time sync off, they drift like mad and the system log is filled with kernel warnings about time changes.
Not at all. I think that kind of thing highlights where Microsoft's priorities are with Windows, and therefore the kinds of things you should be doing with it.
If you buy a civic to take to the racetrack, you shouldn't be amazed it doesn't come with a roll cage. Civics are supposed to be on racetracks, and windows isn't supposed to be a serious-workload OS. It's no problem though, you have plenty of alternatives for both.
44 comments
[ 4.0 ms ] story [ 83.3 ms ] threadThis is cool because it is the best in both dimensions.
Telling the time is hard.
The oscillators on your motherboard, with very basic temperature correction, very basic calibration and only very basic quality control don't tend to do very well on any of these measures. They tend to do better on the high-frequency measures of clock badness, and worse on low-frequency measures, but they aren't stellar in any area. The end-to-end time system, including NTP and software-level drift correction, does an OK job in some circumstances. It does a horrible job in others.
If you go through life assuming your computer's local clock isn't literally out to get you, you're going to be disappointed.
The salient feature here is that quanta are discrete and a transition from one to another is a jump rather than a smooth or continuous shift.
"Quantum leap" connotes an unpredictability in the magnitude of the change in common usage. Over time, the understanding has drifted a bit to include, usually, the idea of significant magnitude.
http://www.merriam-webster.com/dictionary/quantum%20leap
In this particular case it is a perfect use of the phrase, as the advance in accuracy is based on measurements of the atomic electron transitions, also known as quantum leaps.
Also, would this require the clocks and mass to be far away from other large masses like the earth?
[1]: http://www.wolframalpha.com/input/?i=%281+second+%2F430+tril...
http://eu.mio.com/en_gb/global-positioning-system_gps-accura...
[1] http://en.wikipedia.org/wiki/Pulsar_clock
Special and general relativity steam-roll over synched digit accuracy, at introducing any object's mass, any object's displacement.
Not gain or lose over five billion years?
How many digits this time?
Tides, quakes, rain on the roof, defeat the possibility of unbounded synched accuracy.
By definition there would be no more accurate timing device to benchmark it against so is the accuracy cited in the article only theoretical?
The lattice clock in this story has numerous clusters of atoms. They could fill only a few clusters and measure the performance relative to a good clock, then statistically derive how much it would improve by using the full set of clusters.
On a science note, why isn't this a gravity wave detector anyway?
It turns out there are some ultraviolet nuclear transitions. The line widths promise to be obscenely narrow. If they can get it working in a clock, they will be able to directly measure gravitational time dilation of small masses.
http://www.thorium.at/?page_id=4
I could not find many references, but I think clocks mainly make the instrument cheaper, or possible in the case of long-baseline satellite instruments.
Microsoft's Win32 time is only meant to prevent Kerberos error, so under 5 minutes is "fine".
If you buy a civic to take to the racetrack, you shouldn't be amazed it doesn't come with a roll cage. Civics are supposed to be on racetracks, and windows isn't supposed to be a serious-workload OS. It's no problem though, you have plenty of alternatives for both.