Can a clock nerd please ELI5 how this chip differs from a frequency standard like even the old Agilent (then Symmetricon) 5071A, not to speak of more modern variants? For bog-standard businesses that want a on-prem physical backup time standard for IT in case NTP and GPS are not available, would even this chip be overkill?
It's cheap, small, and low-power. It's not a metrology standard like the 5071A. Something like this might be used as a more stable reference oscillator for a radio transceiver. Cell towers typically use rubidium frequency standards, this would likely be part of a package to replace those. They'll also likely show up in high-end test equipment (oscilloscopes, spectrum analyzers, etc).
Note that this thing doesn't provide date and time. It provides a 10MHz square wave output and a 1 pulse per second output. You could use it to make a real time clock, but that's not the intended purpose.
A CSAC is a c(a)esium standard, but it's not a primary cesium standard. A primary standard is required to deliver its rated accuracy without calibration, but the CSAC is more like a rubidium gas cell than a traditional cesium-beam standard. It's very stable relative to most clocks, but it still drifts under the influence of temperature, aging, barometric pressure, and who knows what else.
When you ask "what time is it?", the answer ultimately comes down to an ensemble of cesium-beam clocks, mostly but not exclusively HP/Agilent/Symmetricom/Microsemi 5071As. A 5071A will be about 10x as stable as a CSAC over intervals from seconds to days.
I'm not a clock pro or anything, but it looks like this cites stability numbers around 1e-10 or 1e-11. 5071A says things like 1e-14, but I don't know if it's apples-to-apples. For "portable" applications, NASA's working on the Deep Space Atomic Clock (DSAC) for easier/better space navigation. DSAC is cited as having 1 microsecond per decade stability, or about 3e-15.
While it might be orders of magnitude better than a stabilized quartz oscillator, it looks orders of magnitude away from traditional atomic clocks.
The important thing here is 'chip scale'. This is a device that can be integrated into relatively small packages, enabling a number of use cases.
To give an example - direction finding of the source of a radio signal using multiple drones. This requires the drones to have extremely good time synchronisation so that the time of arrival at each of them can be compared. With a CS atomic package that can be a trivial exercise, without it you need to rely on an external timing reference of some form (e.g. GPS).
I suspect compared to rtk its not really worth the hassle. GPS has been specifically designed to not require a highly disciplined clock for the receiver.
The dev kit looks ridiculously overpriced, but this seems to be the norm even with other electronics --- the "evaluation board", which is often not much more than the device itself mounted on a PCB with some other cheap passives, costs more than 10-100x the actual part.
In this case, however, I couldn't find anywhere selling the actual clock module from a quick search, but there are mentions that it's $1500.
This has existed for a little while now, but doesn't appear to be available in volumes greater than 'sample' (it is not uncommon for semi-conductor manufacturers to jump the gun a lot on something being actually available).
I haven't tried to order one in a month or two though, so that may have changed.
Yeah, even if you have atomic clocks in every datacenter, you need an external reference to synchronize them with each other.
(If you try to calculate the offsets between two clocks by using something like NTP to measure round-trip times over the internet, you're implicitly assuming that packet delays in both directions are symmetrical.)
> Relativity is not just some abstract mathematical theory: understanding it is absolutely essential for our global navigation system to work properly!
I wonder if we could get away by just building it and accounting for the error using some multiplier, without understanding the concept of relativity. I.e., how we would build it if we invented satellites before relativity.
I wonder how well Google's database would work in an interplanetary setting, where due to high relative velocities the property of simultaneity does not exist.
