To keep things in perspective, a vertical shift of a few centimeters could be measured if two of these clocks were placed next to each other, just by the lesser gravity/time dilation at the increased "altitude".
It's an amazing time to be alive. While not this precise, you can have atomic cesium beam clocks of your own for a few thousand dollars each, and some elbow grease.
This is compared to the ~1 mile vertical shift resolution of cesium clocks. The fun part of cesium clocks is that you throw three in the back of a minivan and take them camping!
My calculations says that moving 1cm up or down earths gravity well (at the surface) changes the acceleration of gravity about 5x more than the acceleration you'd feel from a 100kg mass 1m away.
Assuming my math is correct, it's already affected by nearby human scale masses, for certain values of "near".
Huh, I was thinking that it's accelerated gravitational frames that cause the dilation, and I've encountered a lot of statements that argue the same. This is from wikipedia: "This is because gravitational time dilation is manifested in accelerated frames of reference or, by virtue of the equivalence principle, in the gravitational field of massive objects."
However, according to that logic, an object located in a cavity in the center of earth should experience no more dilation than an object outside the earth's potential well, because the gravitational forces / curvature gradient cancels out, and should be zero. But that isn't the case according to the same sources, for example, Wikipedia says' "Relative to Earth's age in billions of years, Earth's core is in effect 2.5 years younger than its surface."
Something's not right about how we verbalize this story about gravity
It's a very subtle point! The trick here is that you need to be careful when talking about the reference frames.
To an observer at the infinity, a clock at the core of the Earth will tick slower than a clock on the surface of the Earth because the "core clock" is sitting in a more curved space, and that's it.
The difference between the clock on the surface of the Earth and the clock at the core is that the surface clock can't follow the "straight lines" (geodesics) in that curved space. So it experiences acceleration due to the force of inertia. And the thing preventing that movement is the repulsive force between atoms that make up the bulk of the Earth.
If this repulsive force magically disappears, then the Earth's atoms will immediately start moving at the straight lines, in trajectories that will lead them all into a point at the center of the Earth.
To add: the force of inertia due to moving in curved lines instead of geodesics depends on the "steepness" of the curved space. Which decreases as you reach the center of the Earth. So you get essentially the same result as with the classic Newtonian gravity, but through an entirely different path.
From what I understand, the relativistic effects are a lot less sensitive to nearby mass than the acceleration due to gravity (basically a 1/r relationship instead of 1/r^2). So while a sensitive enough gravimeter can pick up a nearby fairly heavy mass moving and such an elevation change, an atomic clock is going to be much more sensitive to elevation changes than nearby changes in density.
There's an article, I think on wired.com, years ago about exactly this. It talked about using a vast array of accurate clocks as a kind of radar. Seems plausible only with a few more orders of magnitude accuracy and miniaturization.
I have had to talk myself out of the buying one of those atomic clocks on a chip. I'm kind of a watch geek, and I find the entire history of timekeeping to be endlessly fascinating, and I think it would be so cool to have atomic clock level of precision just running in my house. I've just never been able to justify actually buying one; outside of looking at a console and saying "how cool it that!?!", I don't know what I'd do with it.
I'm not above buying a toy to look at logs and geek out over it, but I can't justifying spending several grand for it.
I fear this altimeter idea may be scuppered by local variations in the Earth’s density (it’s not an exactly uniform sphere of rock). Or maybe that just means the clocks could be great density-mappers!
On second thought, you need a base station on the ground to tell you its time for comparison anyway, so if that base station is nearby the density thing should mostly work itself out
> It's an amazing time to be alive. While not this precise, you can have atomic cesium beam clocks of your own for a few thousand dollars each, and some elbow grease.
How hard or expensive would it be for a reasonably equipped lab to build their own optical clock though? I see there are optical clocks the size of few rack units on the market for a rather hefty price, are the materials needed that expensive or is it just the expertise?
The lasers alone set you back many tens of k so it's not really possible to do on the cheap presently, even if a lot of the cost is expertise and the high overhead of R&D costs when only producing small number of units.
Oh and to know if it's any good you have to either build two (ideally more) of them to compare against each other (ideally using different approaches so their errors are less correlated), or have access to a clock better than the one you're building to compare to. So you can rarely get away with building just one if you want to know if you've succeeded.
What keeps your average home experimenter from building an optical clock is the fact that femtosecond combs are still way too expensive and exotic. Some progress has been made -- you can get them from ThorLabs, for instance ( https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=11... ) -- but they are still in the "Call for pricing and lead time" category.
Once optical comb sources are commoditized to the extent that solid-state lasers are now, a lot of fun stuff will become possible.
Heh, I would not ever have expected to see my company mentioned on HN. I'm the software tech lead at Menlo Systems, we're building those frequency combs that Thorlabs sells.
Re the commoditization: Part of the problem is that customers, especially the scientific ones, don't want "commodity" frequency combs. Nearly every comb we sell is tailored to the specific customer in one way or another.
Industrial customers start to be interested in frequency combs more and more. I guess this will be the clientele that values off-the-shelf products more, eventually paving the way for commoditization.
The presence of individuals such as yourself is what makes the HN comments such a frequently meaningful place to find insightful discussions. Thanks for the context!!
In a 2010 experiment based on an older version of this clock[0], NIST succeeding in measuring the gravitational time dilation across a 33 cm vertical separation—a frequency difference of 4.1×10^{-17}—with 140,000 seconds of integration time (<2 days). I don't really understand how that worked.
Time dilation from general relativity is approximately gh/c^2 (1e-18 -ish), which is an order of magnitude bigger than the uncertainty on your clock frequency (1e-19 -ish).
But you would need a more precise characterization of the clock to answer this.
Edit: If you just have clock output in ticks, you also need enought time to elapse to get a deviation of at least one tick between both bot clocks you are comparing. This is a big limitation, because at a clock rate of 1GHz you are still waiting for like 30 years (!!). (In practice you could probably cheat a bit to get around this limit)
These are optical clocks, so their rates are in the hundreds of THz to PHz range (1.121 PHz in the case of the Al+ clock). A 1e-18 shift corresponds to a ~mHz frequency deviation which a decent frequency counter can resolve in around 1s. However, no optical clock is actually that stable at short time scales, and the averaging is required to get rid of these fluctuations.
>Edit: If you just have clock output in ticks, you also need enought time to elapse to get a deviation of at least one tick between both bot clocks you are comparing. This is a big limitation, because at a clock rate of 1GHz you are still waiting for like 30 years (!!). (In practice you could probably cheat a bit to get around this limit)
In practice with this level of precision you are usually measuring the relative phase of the two clocks, which allows substantially greater resolution than just looking at whole cycles, which is 'cheating' to some degree, I guess. (The limit is usually how noisy your phase measurement is)
(To give some intuition, imaging comparing two pendulum clocks. I think you can probably see how if you take a series of pictures of the pendulums next to each other you could gauge whether one of them is running fast relative to the other, and by how much, without one completing one full swing more than the other)
Instantly more or less. Time instantly moves differently at altitude because you are in a weaker gravitational field. The time dilation effect would be noticeable after 1 (or at most a few) ticks of the clocks.
I'm very skeptical of this claim. While the physical effect of time dilation acts immediately, I expect it would take many many ticks of both clocks before the rate difference between them became resolvable.
I don't understand. Wouldn't it only be possible to find out by comparing two identical clocks that were at different altitudes for some larger number of ticks, allowing you to then compare the elapsed ticks?
How would you conduct such an experiment?
My mental model is that I have a black box that outputs an electrical signal every tick, and then maybe we could just figure out which clock ticked first with a simple circuit. But that seems like we would need to sync them, and that it's fundamentally wrong due to the fact that the information of the tick is also subject to the speed of light. I don't know much beyond high school physics, fwiw.
My comment here might give some intuition for it: https://news.ycombinator.com/item?id=44576004 . You do need to measure for some time, because the measurement of the clocks with respect to each other is noisy, but you don't need to wait for there to be a whole 'tick' of extra time between them.
Yes, and no. The time-dilation effect will happen instantly, but the more quickly you want to observe it, the better your measurement's S/N ratio will have to be... and that, in turn, requires narrow measurement bandwidths that imply longer observation times.
So then the question has to be asked, does the effect really happen instantly? Or do the same mechanisms that impose an inverse relationship between bandwidth and SNR mean that, in fact, it doesn't happen instantly at all?
According to ChatGPT, the speedup factor for getting 10 cm higher is 1 + 1.09e−17. (With ΔT = gh /(c^2) The math seems to check out, but not sure if the formula itself is correct.) Surely, if the clock ticks at rate 1e-19 in a second, i.e. one tick is hundred times smaller than the dilation difference in a second, the clock would still need at least a hundreth of a second to accumulate enough ticks for the count of ticks to differ even by one tick because of the dilation.
The frequency that is actually counted with a digital counter in this clock is only 500 MHz (i.e. after a frequency divider, because no counter can be used at the hundreds of THz of an optical signal).
Nevertheless, in order to measure a frequency difference between two optical clocks you do not need to count their signals. The optical signals can be mixed in a non-linear optical medium, which will provide a signal whose frequency is equal to the difference between the input frequencies.
That signal might have a frequency no greater than 1 GHz, so it might be easy to count with a digital counter.
Of course, the smaller the frequency difference is, the longer must be the time used for counting, to get enough significant digits.
The laser used in this clock has a frequency around 200 THz (like for optical fiber lasers), i.e. about 2E14 Hz. This choice of frequency allows the use of standard optical fibers to compare the frequencies of different optical clocks, even when they are located at great distances.
Mixing the light beams of 2 such lasers, in the case of a 1E-17 frequency difference would give a difference signal with a period of many minutes, which might need to be counted for several days to give an acceptable precision. The time can be reduced by a small factor selecting some harmonic, but it would still be of some days.
Let's imagine that there is a huge amount of time dilation (we live on the surface of a neuron star or something). By climbing a bit, we experience 1.1 seconds instead of 1.0 seconds experienced by someone who left down.
We have a clock that can measure milliseconds as the smallest tick. But climbing up, back down, and comparing the amount of ticks won't let us conclude anything after a single millisecond. If anything, we must spend at least 11 milliseconds up to have a noticeable 11 to 10 millisecond difference.
Now, if the dilation was 1.01 seconds vs 1.00, we would need to spend at least 101 milliseconds up, to get a minimal comparison between 101 and 100 milliseconds.
Thinking in terms of 'ticks' over-discretises it. You can in practice measure frequency to much more precision than any discrete cycle time in the system you're using, because you can usually measure phase, i.e. you're not just seeing some on-off flash, you're seeing a gradual pulse or wave, and it's how accurately you can measure that pulse (in terms of SNR) which sets your integration time for a given precision, not how rapidly you're measuring it.
They would have to be extremely low-frequency. Plus, you'd need something to compare them to that wasn't affected by the wave, which is hard. (You'd need as accurate a clock, placed at a distance greater than the gravitational wavelength.)
I don't think this changes the way the second is defined. Rather, that definition describes some theoretical ideal where the cesium transitions are all perfectly equally spaced over the course of the second.
I think this new clock is simply able to generate more precisely spaced ticks than those of a traditional Cs clock. Less jitter and variation in the timing of those ticks. Similar to how a one-hour water clock or sand timer's runtime will vary between "transitions", but a one-hour quartz stopwatch timer is much more regular. I could keep going, but I'm already out on a limb so I'll stop before my own uncertainty rises too much.
(Edit: I read the article. I don't think my words above are correct.)
I left a comment on the first that summarizes the second one, which describes how they're working on a new type of atomic "nuclear" clock based on the atomic nucleus instead of electron orbitals. It doesn't mention the accuracy, I wonder how it would compare to this "ion" clock.
The largest uncertainty of the Al+ clock is caused by relativistic time dilation due to the ion moving around a bit in the ion trap. This would affect a 229Th clock in the same way, but maybe the larger atomic mass would help.
as a layman, wouldn't you need a more-accurate clock to measure the accuracy of a clock? How is clock accuracy measured when the clock is the most accurate clock?
You define it to be accurate, and then measure the precision.
You can build two and see how much they shift relative to each other. That gives you precision.
So what's the point of a clock if you just define it to be correct? Again, having two clocks is what makes it interesting. Some people have commented that according to general relatively there will be measuralbe time dilation, but there are other fun experiments, e.g.
- Measure shift of fundamental "constants": If you have two clocks that use different elements, the frequency ratio can be related to some things we thought were constants in the universe. If they shift, they aren't constant.
- Look for preferred directions in space: does one clock give a different reading if you turn it on its side?
- Some theories predict that dark matter might induce a frequency shift in these clocks. Put the clocks far apart and look for spacial modulations in the dark matter density.
- Measure anything else that had to be tweaked to make the clock stable. This includes the magnetic field, for example, so the clock is also a really sensitive magnetometer.
This is a great question that always pops up when precision clocks are discussed.
Consider the construction of a precision clock, then build 2 (or more) of them. Take a pair set them to the same initial time, and then let them count time (same height, etc).
Ideally, a pair of perfect clocks would display the same time over time. In practice you see the clocks slowly walk off with respect to each other (a systematic component and a random component), a reasonable first order approximation is to pretend the difference in displayed time shows a random walk behavior. A collection of a large number of clocks would behave like a collection of random walk instances, diffusing in delta-time space.
A poor clock construction would diffuse more quickly, and a better clock construction would diffuse more slowly.
One doesn't need a perfect clock reference to measure the random walk of a clock type / construction. Just compare 2 identically constructed and used ones.
This new clock is way more accurate than that baseline definition though. The SI definition is a practical one for almost everything, the new clock is useful if you are e.g. looking for shifts in physical constants over time.
You are not actually measuring the accuracy of the clock, rather you are measuring the magnitude of the noise. The clock source itself is something physically fundamental and unchanging, but it gets mixed with noise.
For example even very small magnetic fields will change the clock speed, thermal changes will as well (so will lots of other things). So you try to shield from that, and keep the temperature stable (and of course you need to figure out every other things that could add noise).
Then you measure all those influences that you just are unable to control, and calculate what affect they have on the clock, and that's your accuracy number.
One way to directly measure that, instead of calculate it, is to have two identical clocks, synchronize them, and let them run. Then compare them, and see if they differ. (Watch out for relativity messing with time.)
You define time based on a physical phenomenon which does not vary.
For example, every electron is exactly the same as every other electron, they do not vary in the slightest. You utilize properties like that to make exact references to time.
If this were true, then why even bother to make atomic clocks? Why would an article about the "most accurate clock" be interesting to smart people like HN readers if there's no objective measure of accuracy (or if it didn't matter)? The correct answer is in a sibling comment to yours: we base it on other things we know (or believe, anyway) to be constant.
I'm fully aware, but you seem to have misinterpreted what I was saying.
If "all the clocks are wrong" it doesn't matter as long as they are consistent. (in the case of atomic clocks, frequency of energy transitions within atoms)
All ntp servers get the average of atomic clocks, which is then distributed to all phones and computers.
If the constants from these atomic clocks "are a little bit wrong" it does not matter (for most human activities)
That's why we average them and distribute the average.
For physics related research, this new clock being more precise does have use, but for pretty much everything else, whatever constant we have is good enough as long as it's consistently used.
Back in the day it was someone just running around with a pocket watch giving everyone the time from the clock tower which was calibrated from a sundial and that was good enough.
Replace the sun's shadow with electron transitions and the timekeepers with ntp servers and that's what you have today.
Optical atomic clocks based on trapped single ions like this, and also those based on lattices of neutral atoms do not provide a continuous clock signal.
They are used together with a laser (which is a component included in a so-called frequency comb, which acts as a frequency divider between the hundreds of THz of the optical signal and some hundreds of MHz or a few GHz of a clock signal that can be counted with a digital counter; that digital counter could be used as a date and time clock, except that you would need more such optical clocks, to guard against downtime; the present optical clocks do not succeed to operate for very long times before needing a reset because the trapped ion has been lost from the trap or the neutral atoms have been lost from the optical lattice; therefore you need many of them to implement a continuous time scale).
The laser is the one that provides a continuous signal. In this case the laser produces infrared light in the same band as the lasers used for optical fiber communications, and it is based on glass doped with erbium and ytterbium. The frequency of the laser is adjusted to match some resonance frequency of the trapped ion (in this case a submultiple of the frequency, because the frequency of the transition used in the aluminum ion is very high, in ultraviolet). For very short time intervals, when it cannot follow the reference frequency, because that must be filtered of noise, the stability of the laser frequency is determined by a resonant cavity made of silicon (which is transparent for the infrared light of the laser), which is cooled at a very low temperature, in order to improve its quality factor.
So this is similar to the behavior of the clock of a computer, which for long time intervals has the stability of the clocks used by the NTP servers used by it for synchronization, but for short time intervals it has the stability of its internal quartz oscillator.
This new optical atomic clock has the lowest ever uncertainty for the value of its reference frequency, but being a trapped single ion clock it has a higher noise than the clocks based on lattices of neutral atoms (because those can use thousands of atoms instead of one ion), so its output signal must be averaged over long times (e.g. many days) in order to reach the advertised accuracy.
For short averaging times, e.g. of one second, its accuracy is about a thousand times worse than the best attainable (however, its best accuracy is so high that even when averaged for a few seconds it is about as good as the best microwave clocks based on cesium or hydrogen).
We can both agree that the Big Bang happened 13.8 billion years ago but that's all, we'll still disagree about the timing of everything else. Not even the CMB can be used as an universal rest frame. I'm not a cosmologist though.
If I'm not mistaken this improves a critical bottleneck on GPS precision, solutions to which will open up amazing applications. Driving lane boundaries just being one example.
I'm not sure using a web of satellite borne clocks to do location to say three inches or 10cm accuracy is a great idea, especially when the "device" is moving at say 70mph. Not only that, you will need some amazingly exact maps that agree with the satellites' idea of "location".
Lane awareness accuracy needs some serious looking at. Lane markings are designed for the human eye for obvious reasons. Those white lines and "cats eyes" are quite literally (lol) visual cues to keep you on track. If we want machines to drive for us then we will need other ways to do lane compliance, rather than just assist.
My Chinese designed MG4 EV has lane assist. It certainly has four optical cameras but beyond that, I don't know - I think it has a fore and aft proximity sensor of some sort. I do know it makes some dreadful errors of judgment with regards lanes. It's normally fine on a motorway or dual carriageway but gets confused on even a major A road (eg A30) when there are striations in the road surface or its wet or the sun is at the right angle to make skid marks look like lines.
If roads had something like a steel wire armoured cable buried underneath each lane and boundary line and there was a suitably cheap detector and sensors that could look far enough ahead and model the road then that might be a reasonably cheap option to augment optical systems.
Now, more sensors cost money, sorry, cost profits. They also add complexity and so on.
Sadly I think we will continue to see daft things like ... I gather that Teslas will only sport optical sensors and no LIDAR.
Autonomous cars are possible but they do need to be given a decent amount of, and variety of, ... sensors.
GPS will get you very well located on a map but you need local sensors to sort out localisation issues.
The main errors in GPS are usually due to variation in the propagation of the signals through the atmosphere, and uncertainty in the exact position of the satellites. (Solutions for these exist, but are generally too expensive and/or impractical for a lot of applications. But e.g. realtime GPS on the order of cm is something you can get now, if you pay for a subscription to the right service or operate your own reference base station)
It could improve accuracy and does help a significant error factor, but the biggest error sources in GNSS aren't clock errors. Even a perfect clock could still have several meters of error.
Busy week after our AC died, but I'm taking some notes.
I'm up to three running GPS clocks, one with OXCO, one with TXCO, and one with Rubidium holdover... going to finally be able to say what time it might be more intelligently :)
Also, I hope to meet CuriousMarc this weekend! He's done some fascinating things with Cesium clocks.
Somewhat topical: in case you want authenticated access to NIST NTP servers, you need to send a letter to NIST using the US mail or FAX machine (e-mail is not acceptable).
NIST will reply with a key number and a key value. The reply will be by US mail only, e-mail will never be used.
The office that normally receives US mail and FAX messages currently has limited access, which may result in significant delays in processing requests
I was thinking that it may be the point. This is a service paid by Americans, for Americans, and I suspect they don't want to make it too convenient, especially for non-Americans.
From the link:
> The service will be provided at no charge, and user keys may be used to connect to any of the servers whose addresses are listed below. Additional hardware will be added in the future if the demand for the service is sufficiently great to warrant it.
Making it clear that they are going to shoulder the extra costs.
The article is very good and has some cool pictures of the device. Aluminum is apparently just better than cesium but harder to use and now they have solved the problems preventing it from being the standard.
A single measurement cannot be precise. Precision is a measure of how close multiple measurements are to one another. Accuracy is how close a single measurement is to its true value.
A clock that can measure a point in time to 19 decimal places with respect to its true value is accurate.
Unfortunately, I am not a time lord, so I don't know. But I do know the definitions of these words, and that is what they are. You are free to argue with someone else about the true meaning of time.
A single measurement can _never_ be precise, it is simply not possible.
That's great and all, but stepping back a bit: the US (and the West?) seems to be falling behind in distribution and resiliency. For high-accuracy timing stuff, China has space (BeiDou), terrestrial broadcast (eLoran), and fiber in production:
To keep things in perspective, a vertical shift of a few centimeters could be measured if two of these clocks were placed next to each other, just by the lesser gravity/time dilation at the increased "altitude".
It's an amazing time to be alive. While not this precise, you can have atomic cesium beam clocks of your own for a few thousand dollars each, and some elbow grease.
This is compared to the ~1 mile vertical shift resolution of cesium clocks. The fun part of cesium clocks is that you throw three in the back of a minivan and take them camping!
http://leapsecond.com/great2005/
Could we realistically get so accurate that we can measure time changes due to (human scale) mass movements near a clock?
My calculations says that moving 1cm up or down earths gravity well (at the surface) changes the acceleration of gravity about 5x more than the acceleration you'd feel from a 100kg mass 1m away.
Assuming my math is correct, it's already affected by nearby human scale masses, for certain values of "near".
I believe the time dilation is caused by differences in gravitational potential, not gravitational acceleration, so it would be even worse than that.
Huh, I was thinking that it's accelerated gravitational frames that cause the dilation, and I've encountered a lot of statements that argue the same. This is from wikipedia: "This is because gravitational time dilation is manifested in accelerated frames of reference or, by virtue of the equivalence principle, in the gravitational field of massive objects."
However, according to that logic, an object located in a cavity in the center of earth should experience no more dilation than an object outside the earth's potential well, because the gravitational forces / curvature gradient cancels out, and should be zero. But that isn't the case according to the same sources, for example, Wikipedia says' "Relative to Earth's age in billions of years, Earth's core is in effect 2.5 years younger than its surface."
Something's not right about how we verbalize this story about gravity
It's a very subtle point! The trick here is that you need to be careful when talking about the reference frames.
To an observer at the infinity, a clock at the core of the Earth will tick slower than a clock on the surface of the Earth because the "core clock" is sitting in a more curved space, and that's it.
The difference between the clock on the surface of the Earth and the clock at the core is that the surface clock can't follow the "straight lines" (geodesics) in that curved space. So it experiences acceleration due to the force of inertia. And the thing preventing that movement is the repulsive force between atoms that make up the bulk of the Earth.
If this repulsive force magically disappears, then the Earth's atoms will immediately start moving at the straight lines, in trajectories that will lead them all into a point at the center of the Earth.
To add: the force of inertia due to moving in curved lines instead of geodesics depends on the "steepness" of the curved space. Which decreases as you reach the center of the Earth. So you get essentially the same result as with the classic Newtonian gravity, but through an entirely different path.
A clock at the center of the planet should experience no net force by the mass of the planet.
That would be an amazing proximity sensor. "Looks like time slowed down again, there must be someone close by."
Submarine detection would be interesting with a few such precise clocks.
Given that the mass of a submarine is ~the same as the water it displaces, this wouldn't actually work.
I'm curious if any sci-fi authors were knowledgable and prescient enough to write this into their world building?
If not, it'd make for a pretty cool plot device if done well.
From what I understand, the relativistic effects are a lot less sensitive to nearby mass than the acceleration due to gravity (basically a 1/r relationship instead of 1/r^2). So while a sensitive enough gravimeter can pick up a nearby fairly heavy mass moving and such an elevation change, an atomic clock is going to be much more sensitive to elevation changes than nearby changes in density.
There's an article, I think on wired.com, years ago about exactly this. It talked about using a vast array of accurate clocks as a kind of radar. Seems plausible only with a few more orders of magnitude accuracy and miniaturization.
Asked this and related questions to o3. I do not vouch for the answers at all but you may find it interesting. https://chatgpt.com/share/6876cdd1-dfbc-8011-a55f-6915a90275...
I'd love to hear what the kids remember about this trip. It's been awhile!
I have had to talk myself out of the buying one of those atomic clocks on a chip. I'm kind of a watch geek, and I find the entire history of timekeeping to be endlessly fascinating, and I think it would be so cool to have atomic clock level of precision just running in my house. I've just never been able to justify actually buying one; outside of looking at a console and saying "how cool it that!?!", I don't know what I'd do with it.
I'm not above buying a toy to look at logs and geek out over it, but I can't justifying spending several grand for it.
I was in the same boat some time ago, ended up settling for a Casio radio controlled "atomic" clock for multiple orders of magnitude cheaper
This is a great way to compare ultra-precise clocks! Also, I'm looking forward to Einsteinian altimeters everywhere.
I fear this altimeter idea may be scuppered by local variations in the Earth’s density (it’s not an exactly uniform sphere of rock). Or maybe that just means the clocks could be great density-mappers!
It is easier to measure density with gravimeter than measure gravitational force directly.
Satellite, ACES, was launched recently that uses atomic clocks to accurately measure Earth's gravity field.
Are you saying they would be relatively inaccurate?
I had the same thought, but I still want one!
On second thought, you need a base station on the ground to tell you its time for comparison anyway, so if that base station is nearby the density thing should mostly work itself out
> It's an amazing time to be alive. While not this precise, you can have atomic cesium beam clocks of your own for a few thousand dollars each, and some elbow grease.
How hard or expensive would it be for a reasonably equipped lab to build their own optical clock though? I see there are optical clocks the size of few rack units on the market for a rather hefty price, are the materials needed that expensive or is it just the expertise?
The lasers alone set you back many tens of k so it's not really possible to do on the cheap presently, even if a lot of the cost is expertise and the high overhead of R&D costs when only producing small number of units.
Oh and to know if it's any good you have to either build two (ideally more) of them to compare against each other (ideally using different approaches so their errors are less correlated), or have access to a clock better than the one you're building to compare to. So you can rarely get away with building just one if you want to know if you've succeeded.
Source: I work on the software for these portable optical clocks: https://phys.org/news/2025-07-quantum-clocks-accuracy-curren...
What keeps your average home experimenter from building an optical clock is the fact that femtosecond combs are still way too expensive and exotic. Some progress has been made -- you can get them from ThorLabs, for instance ( https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=11... ) -- but they are still in the "Call for pricing and lead time" category.
Once optical comb sources are commoditized to the extent that solid-state lasers are now, a lot of fun stuff will become possible.
Heh, I would not ever have expected to see my company mentioned on HN. I'm the software tech lead at Menlo Systems, we're building those frequency combs that Thorlabs sells.
Re the commoditization: Part of the problem is that customers, especially the scientific ones, don't want "commodity" frequency combs. Nearly every comb we sell is tailored to the specific customer in one way or another.
Industrial customers start to be interested in frequency combs more and more. I guess this will be the clientele that values off-the-shelf products more, eventually paving the way for commoditization.
The presence of individuals such as yourself is what makes the HN comments such a frequently meaningful place to find insightful discussions. Thanks for the context!!
We aren't far at all from on chip combs or pseudo combs, and that will be fine. More for sensors generally, but you can also have your clocks.
Expertise
> ... vertical shift of a few centimeters could be measured
In what amount of time? Not instantly, right?
In a 2010 experiment based on an older version of this clock[0], NIST succeeding in measuring the gravitational time dilation across a 33 cm vertical separation—a frequency difference of 4.1×10^{-17}—with 140,000 seconds of integration time (<2 days). I don't really understand how that worked.
[0] https://sci-hub.se/https://doi.org/10.1126/science.1192720 ("Optical Clocks and Relativity" (2010))
Time dilation from general relativity is approximately gh/c^2 (1e-18 -ish), which is an order of magnitude bigger than the uncertainty on your clock frequency (1e-19 -ish).
But you would need a more precise characterization of the clock to answer this.
There might be significant noise on individual measurements, meaning that you need to take multiples to get precise enough (see https://en.wikipedia.org/wiki/Allan_variance).
Edit: If you just have clock output in ticks, you also need enought time to elapse to get a deviation of at least one tick between both bot clocks you are comparing. This is a big limitation, because at a clock rate of 1GHz you are still waiting for like 30 years (!!). (In practice you could probably cheat a bit to get around this limit)
These are optical clocks, so their rates are in the hundreds of THz to PHz range (1.121 PHz in the case of the Al+ clock). A 1e-18 shift corresponds to a ~mHz frequency deviation which a decent frequency counter can resolve in around 1s. However, no optical clock is actually that stable at short time scales, and the averaging is required to get rid of these fluctuations.
>Edit: If you just have clock output in ticks, you also need enought time to elapse to get a deviation of at least one tick between both bot clocks you are comparing. This is a big limitation, because at a clock rate of 1GHz you are still waiting for like 30 years (!!). (In practice you could probably cheat a bit to get around this limit)
In practice with this level of precision you are usually measuring the relative phase of the two clocks, which allows substantially greater resolution than just looking at whole cycles, which is 'cheating' to some degree, I guess. (The limit is usually how noisy your phase measurement is)
(To give some intuition, imaging comparing two pendulum clocks. I think you can probably see how if you take a series of pictures of the pendulums next to each other you could gauge whether one of them is running fast relative to the other, and by how much, without one completing one full swing more than the other)
From the article:
So no, not instantly.Instantly more or less. Time instantly moves differently at altitude because you are in a weaker gravitational field. The time dilation effect would be noticeable after 1 (or at most a few) ticks of the clocks.
I'm very skeptical of this claim. While the physical effect of time dilation acts immediately, I expect it would take many many ticks of both clocks before the rate difference between them became resolvable.
I don't understand. Wouldn't it only be possible to find out by comparing two identical clocks that were at different altitudes for some larger number of ticks, allowing you to then compare the elapsed ticks? How would you conduct such an experiment? My mental model is that I have a black box that outputs an electrical signal every tick, and then maybe we could just figure out which clock ticked first with a simple circuit. But that seems like we would need to sync them, and that it's fundamentally wrong due to the fact that the information of the tick is also subject to the speed of light. I don't know much beyond high school physics, fwiw.
My comment here might give some intuition for it: https://news.ycombinator.com/item?id=44576004 . You do need to measure for some time, because the measurement of the clocks with respect to each other is noisy, but you don't need to wait for there to be a whole 'tick' of extra time between them.
Yes, and no. The time-dilation effect will happen instantly, but the more quickly you want to observe it, the better your measurement's S/N ratio will have to be... and that, in turn, requires narrow measurement bandwidths that imply longer observation times.
So then the question has to be asked, does the effect really happen instantly? Or do the same mechanisms that impose an inverse relationship between bandwidth and SNR mean that, in fact, it doesn't happen instantly at all?
According to ChatGPT, the speedup factor for getting 10 cm higher is 1 + 1.09e−17. (With ΔT = gh /(c^2) The math seems to check out, but not sure if the formula itself is correct.) Surely, if the clock ticks at rate 1e-19 in a second, i.e. one tick is hundred times smaller than the dilation difference in a second, the clock would still need at least a hundreth of a second to accumulate enough ticks for the count of ticks to differ even by one tick because of the dilation.
The frequency that is actually counted with a digital counter in this clock is only 500 MHz (i.e. after a frequency divider, because no counter can be used at the hundreds of THz of an optical signal).
Nevertheless, in order to measure a frequency difference between two optical clocks you do not need to count their signals. The optical signals can be mixed in a non-linear optical medium, which will provide a signal whose frequency is equal to the difference between the input frequencies.
That signal might have a frequency no greater than 1 GHz, so it might be easy to count with a digital counter.
Of course, the smaller the frequency difference is, the longer must be the time used for counting, to get enough significant digits.
The laser used in this clock has a frequency around 200 THz (like for optical fiber lasers), i.e. about 2E14 Hz. This choice of frequency allows the use of standard optical fibers to compare the frequencies of different optical clocks, even when they are located at great distances.
Mixing the light beams of 2 such lasers, in the case of a 1E-17 frequency difference would give a difference signal with a period of many minutes, which might need to be counted for several days to give an acceptable precision. The time can be reduced by a small factor selecting some harmonic, but it would still be of some days.
To make this even clearer:
Let's imagine that there is a huge amount of time dilation (we live on the surface of a neuron star or something). By climbing a bit, we experience 1.1 seconds instead of 1.0 seconds experienced by someone who left down.
We have a clock that can measure milliseconds as the smallest tick. But climbing up, back down, and comparing the amount of ticks won't let us conclude anything after a single millisecond. If anything, we must spend at least 11 milliseconds up to have a noticeable 11 to 10 millisecond difference.
Now, if the dilation was 1.01 seconds vs 1.00, we would need to spend at least 101 milliseconds up, to get a minimal comparison between 101 and 100 milliseconds.
Thinking in terms of 'ticks' over-discretises it. You can in practice measure frequency to much more precision than any discrete cycle time in the system you're using, because you can usually measure phase, i.e. you're not just seeing some on-off flash, you're seeing a gradual pulse or wave, and it's how accurately you can measure that pulse (in terms of SNR) which sets your integration time for a given precision, not how rapidly you're measuring it.
> Let's imagine that there is a huge amount of time dilation (we live on the surface of a neuron star or something).
That idea is the premise of https://en.wikipedia.org/wiki/Incandescence_(novel)
https://en.wikipedia.org/wiki/Allan_variance
It takes a longer measurement to be more confident.
Could you detect gravity waves with accurate enough clocks?
They would have to be extremely low-frequency. Plus, you'd need something to compare them to that wasn't affected by the wave, which is hard. (You'd need as accurate a clock, placed at a distance greater than the gravitational wavelength.)
I think.
Isn’t the second defined as a specific number of cesium transitions?
How can anything … you know what? Never mind. No matter what answer anyone provides, I won’t understand.
> Isn’t the second defined as a specific number of cesium transitions?
Yes
> How can anything …
So your cesium counting device will fauthfully provide such a count and depending on their altitude it will be at different rates.
Both clocks are each experiencing time at the usual one second per second but gravity dilates spacetime.
Locally, a second is always a second, but from everywhere there is no such asbsolute, just as there is no universal "now".
I don't think this changes the way the second is defined. Rather, that definition describes some theoretical ideal where the cesium transitions are all perfectly equally spaced over the course of the second.
I think this new clock is simply able to generate more precisely spaced ticks than those of a traditional Cs clock. Less jitter and variation in the timing of those ticks. Similar to how a one-hour water clock or sand timer's runtime will vary between "transitions", but a one-hour quartz stopwatch timer is much more regular. I could keep going, but I'm already out on a limb so I'll stop before my own uncertainty rises too much.
(Edit: I read the article. I don't think my words above are correct.)
SKO BUFFS
I briefly worked at NOAA, on this same campus, and I loved walking around NIST. Such a cool building. The entire campus is at risk -> https://www.cpr.org/2025/07/01/proposed-noaa-budget-would-cl...
Some recent discussions about atomic clocks:
New atomic fountain clock joins group that keeps the world on time (nist.gov) | 118 points | 76 days ago | 33 comments | https://news.ycombinator.com/item?id=43831792
Major leap for nuclear clock paves way for ultraprecise timekeeping (nist.gov) | 12 points | 7 months ago | 10 comments | https://news.ycombinator.com/item?id=42362215
I left a comment on the first that summarizes the second one, which describes how they're working on a new type of atomic "nuclear" clock based on the atomic nucleus instead of electron orbitals. It doesn't mention the accuracy, I wonder how it would compare to this "ion" clock.
The largest uncertainty of the Al+ clock is caused by relativistic time dilation due to the ion moving around a bit in the ion trap. This would affect a 229Th clock in the same way, but maybe the larger atomic mass would help.
as a layman, wouldn't you need a more-accurate clock to measure the accuracy of a clock? How is clock accuracy measured when the clock is the most accurate clock?
How do you measure the accuracy of a clock? What if every clock is wrong, just a little bit?
You define it to be accurate, and then measure the precision.
You can build two and see how much they shift relative to each other. That gives you precision.
So what's the point of a clock if you just define it to be correct? Again, having two clocks is what makes it interesting. Some people have commented that according to general relatively there will be measuralbe time dilation, but there are other fun experiments, e.g.
- Measure shift of fundamental "constants": If you have two clocks that use different elements, the frequency ratio can be related to some things we thought were constants in the universe. If they shift, they aren't constant.
- Look for preferred directions in space: does one clock give a different reading if you turn it on its side?
- Some theories predict that dark matter might induce a frequency shift in these clocks. Put the clocks far apart and look for spacial modulations in the dark matter density.
- Measure anything else that had to be tweaked to make the clock stable. This includes the magnetic field, for example, so the clock is also a really sensitive magnetometer.
This is a great question that always pops up when precision clocks are discussed.
Consider the construction of a precision clock, then build 2 (or more) of them. Take a pair set them to the same initial time, and then let them count time (same height, etc).
Ideally, a pair of perfect clocks would display the same time over time. In practice you see the clocks slowly walk off with respect to each other (a systematic component and a random component), a reasonable first order approximation is to pretend the difference in displayed time shows a random walk behavior. A collection of a large number of clocks would behave like a collection of random walk instances, diffusing in delta-time space.
A poor clock construction would diffuse more quickly, and a better clock construction would diffuse more slowly.
One doesn't need a perfect clock reference to measure the random walk of a clock type / construction. Just compare 2 identically constructed and used ones.
Since 1967 the second has a physical defintion:
https://en.wikipedia.org/wiki/Second#Atomic_definition
This new clock is way more accurate than that baseline definition though. The SI definition is a practical one for almost everything, the new clock is useful if you are e.g. looking for shifts in physical constants over time.
You are not actually measuring the accuracy of the clock, rather you are measuring the magnitude of the noise. The clock source itself is something physically fundamental and unchanging, but it gets mixed with noise.
For example even very small magnetic fields will change the clock speed, thermal changes will as well (so will lots of other things). So you try to shield from that, and keep the temperature stable (and of course you need to figure out every other things that could add noise).
Then you measure all those influences that you just are unable to control, and calculate what affect they have on the clock, and that's your accuracy number.
One way to directly measure that, instead of calculate it, is to have two identical clocks, synchronize them, and let them run. Then compare them, and see if they differ. (Watch out for relativity messing with time.)
You define time based on a physical phenomenon which does not vary.
For example, every electron is exactly the same as every other electron, they do not vary in the slightest. You utilize properties like that to make exact references to time.
it doesn't matter? it's the average of multiple atomic clocks and that's the time we distribute.
it's a human construct so whatever is agreed upon is correct.
If this were true, then why even bother to make atomic clocks? Why would an article about the "most accurate clock" be interesting to smart people like HN readers if there's no objective measure of accuracy (or if it didn't matter)? The correct answer is in a sibling comment to yours: we base it on other things we know (or believe, anyway) to be constant.
I'm fully aware, but you seem to have misinterpreted what I was saying.
If "all the clocks are wrong" it doesn't matter as long as they are consistent. (in the case of atomic clocks, frequency of energy transitions within atoms)
All ntp servers get the average of atomic clocks, which is then distributed to all phones and computers.
If the constants from these atomic clocks "are a little bit wrong" it does not matter (for most human activities)
That's why we average them and distribute the average.
For physics related research, this new clock being more precise does have use, but for pretty much everything else, whatever constant we have is good enough as long as it's consistently used.
Back in the day it was someone just running around with a pocket watch giving everyone the time from the clock tower which was calibrated from a sundial and that was good enough.
Replace the sun's shadow with electron transitions and the timekeepers with ntp servers and that's what you have today.
I see. Thanks for the added context; that's much clearer.
Next family meetup. Uncle: “So, how’s work?” Them: “Yeah, I just made the universe’s most ridiculous clock this year.” Absolute power move.
Is it a “clock” or a “clock signal”, in a similar sense that position encoders can be relative or “absolute” (but only within a specific range)?
Optical atomic clocks based on trapped single ions like this, and also those based on lattices of neutral atoms do not provide a continuous clock signal.
They are used together with a laser (which is a component included in a so-called frequency comb, which acts as a frequency divider between the hundreds of THz of the optical signal and some hundreds of MHz or a few GHz of a clock signal that can be counted with a digital counter; that digital counter could be used as a date and time clock, except that you would need more such optical clocks, to guard against downtime; the present optical clocks do not succeed to operate for very long times before needing a reset because the trapped ion has been lost from the trap or the neutral atoms have been lost from the optical lattice; therefore you need many of them to implement a continuous time scale).
The laser is the one that provides a continuous signal. In this case the laser produces infrared light in the same band as the lasers used for optical fiber communications, and it is based on glass doped with erbium and ytterbium. The frequency of the laser is adjusted to match some resonance frequency of the trapped ion (in this case a submultiple of the frequency, because the frequency of the transition used in the aluminum ion is very high, in ultraviolet). For very short time intervals, when it cannot follow the reference frequency, because that must be filtered of noise, the stability of the laser frequency is determined by a resonant cavity made of silicon (which is transparent for the infrared light of the laser), which is cooled at a very low temperature, in order to improve its quality factor.
So this is similar to the behavior of the clock of a computer, which for long time intervals has the stability of the clocks used by the NTP servers used by it for synchronization, but for short time intervals it has the stability of its internal quartz oscillator.
This new optical atomic clock has the lowest ever uncertainty for the value of its reference frequency, but being a trapped single ion clock it has a higher noise than the clocks based on lattices of neutral atoms (because those can use thousands of atoms instead of one ion), so its output signal must be averaged over long times (e.g. many days) in order to reach the advertised accuracy.
For short averaging times, e.g. of one second, its accuracy is about a thousand times worse than the best attainable (however, its best accuracy is so high that even when averaged for a few seconds it is about as good as the best microwave clocks based on cesium or hydrogen).
Thank you for that very excellent reply.
Is there really an absolute reference point to measure time other than the big bang or something?
We can both agree that the Big Bang happened 13.8 billion years ago but that's all, we'll still disagree about the timing of everything else. Not even the CMB can be used as an universal rest frame. I'm not a cosmologist though.
Relativity. Every path starting at the big bang to the present has its own unique clock.
These clocks can measure the difference in the flow of time between your head and your feet (and quite a lot more accurate than that)
The clock signals can be counted and are accurate over long periods, it is not just a rate that drifts.
Being able to count trillions of ticks is entirely possible in clocks or rotary encoders, just nobody bothers to do so on rotary encoders very often.
Preprint,
https://arxiv.org/abs/2504.13071 ("High-Stability Single-Ion Clock with $5.5\times10^{-19}$ Systematic Uncertainty")
If I'm not mistaken this improves a critical bottleneck on GPS precision, solutions to which will open up amazing applications. Driving lane boundaries just being one example.
I'm not sure using a web of satellite borne clocks to do location to say three inches or 10cm accuracy is a great idea, especially when the "device" is moving at say 70mph. Not only that, you will need some amazingly exact maps that agree with the satellites' idea of "location".
Lane awareness accuracy needs some serious looking at. Lane markings are designed for the human eye for obvious reasons. Those white lines and "cats eyes" are quite literally (lol) visual cues to keep you on track. If we want machines to drive for us then we will need other ways to do lane compliance, rather than just assist.
My Chinese designed MG4 EV has lane assist. It certainly has four optical cameras but beyond that, I don't know - I think it has a fore and aft proximity sensor of some sort. I do know it makes some dreadful errors of judgment with regards lanes. It's normally fine on a motorway or dual carriageway but gets confused on even a major A road (eg A30) when there are striations in the road surface or its wet or the sun is at the right angle to make skid marks look like lines.
If roads had something like a steel wire armoured cable buried underneath each lane and boundary line and there was a suitably cheap detector and sensors that could look far enough ahead and model the road then that might be a reasonably cheap option to augment optical systems.
Now, more sensors cost money, sorry, cost profits. They also add complexity and so on.
Sadly I think we will continue to see daft things like ... I gather that Teslas will only sport optical sensors and no LIDAR.
Autonomous cars are possible but they do need to be given a decent amount of, and variety of, ... sensors.
GPS will get you very well located on a map but you need local sensors to sort out localisation issues.
The main errors in GPS are usually due to variation in the propagation of the signals through the atmosphere, and uncertainty in the exact position of the satellites. (Solutions for these exist, but are generally too expensive and/or impractical for a lot of applications. But e.g. realtime GPS on the order of cm is something you can get now, if you pay for a subscription to the right service or operate your own reference base station)
It could improve accuracy and does help a significant error factor, but the biggest error sources in GNSS aren't clock errors. Even a perfect clock could still have several meters of error.
Someone call Hodinkee to write about this.
Jeff Geerling
I'm not a fan, I'm a human being.
I honestly just Ctrl+F'd for him to see if he commented yet.
Busy week after our AC died, but I'm taking some notes.
I'm up to three running GPS clocks, one with OXCO, one with TXCO, and one with Rubidium holdover... going to finally be able to say what time it might be more intelligently :)
Also, I hope to meet CuriousMarc this weekend! He's done some fascinating things with Cesium clocks.
Somewhat topical: in case you want authenticated access to NIST NTP servers, you need to send a letter to NIST using the US mail or FAX machine (e-mail is not acceptable).
NIST will reply with a key number and a key value. The reply will be by US mail only, e-mail will never be used.
The office that normally receives US mail and FAX messages currently has limited access, which may result in significant delays in processing requests
https://www.nist.gov/pml/time-and-frequency-division/time-se...
(things you discover when you implement fedramp)
I wonder if they'd consider implementing NTS.
https://github.com/jauderho/nts-servers/tree/main
NTS uses AES-SIV and looks like it's not NIST approved and doesn't have FIPS validated implementation.
in other words - no
Seems pretty tight for those outside of the USA. Is FAX allowed for those who live outside?
I was thinking that it may be the point. This is a service paid by Americans, for Americans, and I suspect they don't want to make it too convenient, especially for non-Americans.
From the link:
> The service will be provided at no charge, and user keys may be used to connect to any of the servers whose addresses are listed below. Additional hardware will be added in the future if the demand for the service is sufficiently great to warrant it.
Making it clear that they are going to shoulder the extra costs.
better question is will be they willing to mail you reply outside of usa.
if you really want it, there are plenty of services that provide you with virtual mailbox in usa
The article is very good and has some cool pictures of the device. Aluminum is apparently just better than cesium but harder to use and now they have solved the problems preventing it from being the standard.
Does it come with a snooze button?
Pedant here!
> NIST researchers have made the most accurate atomic clock to date — one that can measure time down to the 19th decimal place.
That's precision, not accuracy.
Nope, it is not.
A single measurement cannot be precise. Precision is a measure of how close multiple measurements are to one another. Accuracy is how close a single measurement is to its true value.
A clock that can measure a point in time to 19 decimal places with respect to its true value is accurate.
What is that true value? And was it accurate?
Unfortunately, I am not a time lord, so I don't know. But I do know the definitions of these words, and that is what they are. You are free to argue with someone else about the true meaning of time.
A single measurement can _never_ be precise, it is simply not possible.
Tell me you have never shot a rifle for score, without saying you haven't.
You can be accurate, precise, or both.
Said clock may be precise, but not accurate.
It’s about time.
Yet another advance brought to you by US tax dollars funding basic science for the public good.
Please don't let them embezzle the future of scientific innovation.
This is the science that needs to stay. The crap that pretends to be science can get gutted.
Imagine trying to set your oven and microwave clock to line up with it precisely after they reset from a power outage
That's great and all, but stepping back a bit: the US (and the West?) seems to be falling behind in distribution and resiliency. For high-accuracy timing stuff, China has space (BeiDou), terrestrial broadcast (eLoran), and fiber in production:
* https://rntfnd.org/2024/10/03/china-completes-national-elora...
As we've seen regularly, GPS/GNSS has major risks with it, and it seems to have become a single point of failure:
* https://gpsjam.org
* https://www.marineinsight.com/shipping-news/msc-container-sh...
* https://gcaptain.com/gps-jamming-in-strait-of-hormuz-raises-...
So fun even NIST can't get right the difference between precise and accurate