Single-Ion Clock 100 Times More Accurate Than Atomic Clock 169
New submitter labnet writes with this excerpt from news.com.au: "University of New South Wales School of Physics professor Victor Flambaum has found a method of timekeeping nearly 100 times more accurate than the best atomic clocks. By using the orbit of a neutron around an atomic nucleus he says the system stays accurate to within 1/20th of a second over billions of years. Although perhaps not for daily use, the technology could prove valuable in science experiments where chronological accuracy is paramount, Prof Flambaum said."
Re:Orbit around a nucleus? (Score:4, Insightful)
How do you measure how accurate it is? (Score:4, Insightful)
If an atomic clock is your most accurate timepiece then how on earth can you tell if something is more accurate?
Can someone explain?
Also , given that a second is defined in terms of the ceasium atom as used in atomic clocks then surely anything that deviates from this is by definition LESS accurate (if you see what I mean)?
Re:Eventually... (Score:4, Insightful)
Re:How do you measure how accurate it is? (Score:4, Insightful)
The same way it always was. Think of how you'd do it in any sort of mechanical measurements. You don't need the same level of accuracy to determine that something is more accurate. Most measurements have nice properties that must hold when you repeat the measurements, such as linearity. All you have to do, then, is to use the assumedly more accurate device to characterize the errors of a less accurate one. If you can reproduce your results and various expected properties hold, then there's no other explanation but that your new device is in fact more accurate.
The deal with the caesium atom is that it only defines a second to a certain accuracy. If you have a better time reference, it's not by definition less accurate, it's just that your standard has accuracy only to so many decimal digits and when you're past that you must get a better standard. You can use the better reference to characterize the inaccuracies in your standard (say various drifts, phase noise in case of time references, etc). Eventually, you redefine the second using the better standard, and you do it pretty much by appending some arbitrarily chosen digits to the new definition that reproduces the old one. They had second defined however, then they measured it using the caesium clock, got a bunch of results, averaged them, and said: that's the new second. A whole bunch of digits of the new definition were pretty arbitrary -- they original definition wasn't able to provide you with stable digits all the way. Same thing will happen again: the new clock will be used to measure the cesium one, and they'll average things and the new second will be a few orders of mangnitude more cycles of this nuclear clock; it will be matching the old clock within the old clock's accuracy, but the now-added digits will be entirely arbitrary. This is how it has happened with pretty much all the other measurements (distance, weight, etc).
Re:Eventually... (Score:4, Insightful)
a man with three clocks knows if one of his clocks is not working correctly.
So does a man with two clocks. But a man with three clocks may know which one.
Re:How do you measure how accurate it is? (Score:4, Insightful)
If the accuracy is defined as fractions of a second over billion years - how do they know its going to last a billion years
Run the reciprocal and test your frequency. You know that saying about how in europe they think hundreds of miles (err KM) is far away and hundreds of years is recent, but in the US they think hundreds of miles is a daily commute and hundreds of years is ancient? Well billions of seconds is a long time, but billions of cycles per second is actually medium to low frequency in the RF world now a days, depending I guess on industry (that would still be considered kind of fast in the PLC/VFD field, but truly ancient great-grandfatherly stuff in the radar world)
So you've got three atomic clocks (now a days a ebay special Rb clock is about $100 surplus) and use that to drive three sets of ham radio microwave experimenters gear at 10 GHz (which is not cutting edge anymore). Hmm. 10 billion hz. suddenly fractional parts per billion becomes fractional hz which a piano tuner has no real problem detecting.
This isn't exactly how it works, but as a thought experiment you hook up your 10gig ethernet and drive it with this clock and hack the driver for variable length packets... If you think you have better than 0.1 ppb clock, then you should be able to transmit a billion bit packet and not fall out of frame sync (which at 10 gigs only takes a tenth of a second). This is not exactly the modulation method used by real 10gigE and not exactly how you test it, but it within the realm of the general idea.
Good luck doing modern ham radio stuff like bouncing microwave signals off the moon using the more exotic low SNR digital modes without at least PPB level frequency accuracy. Freq stability is a factor at 10 GHz until at least 10e-9 for that kind of work... luckily 10e-11 is cheap and off the (ebay) shelf for $200 or so GPSDO or old Rb oscillators.
100 times more accurate? (Score:4, Insightful)
If Server A has 90% uptime and Server B has 99% uptime, that does not mean that Server B is up 10x more than Server A, even though Server A is down 10x more than Server B. In fact, Server B is only 10% better than Server A. Or, 1/10 as bad.*
So, while the old clock may drift 100x more than this new one in a certain amount of time, or this new one might last 100x longer before drifting a certain amount (or whatever--the .au article is total puff and I don't care enough to look at the source), it is almost certainly not 100x more accurate. At best, it's 1/100th as inaccurate.
* The difference between 36 days of downtime per year versus 4 days might be the difference between "useful" and "completely worthless", making Server B 100x better, but that's not what we're measuring here.