I don't believe it's necessary to have multiple GPS antennas (one per device), unless signal path redundancy is required. A good GPS distribution box like from Time Machines or GPS Source can split the antenna signal to many devices without an issue.
A signal distribution box used from eBay is a lot cheaper than a good outdoor GPS antenna!
Though if you have enough cable and enough antennas already, no harm in having a little array like in OP.
>I don't believe it's necessary to have multiple GPS antennas (one per device), unless signal path redundancy is required
Don't those splitters typically have an amplifier, so become a single point of failure?
If you're going through the trouble to have multiple time servers on your network, you probably want to make sure that an amplifier failure doesn't take them all down at the same time.
GPS antennas typically need power. GPS receivers expect to be the sole power supplier. Simple, cheap signal distribution won't handle this. There are splitters that can allow one downstream port to supply power, but that fails when that single receiver fails. There are other splitters that are active-active. I guess you're claiming that the boxes from Time Machines are such active-active devices.
Those are not cheaper than an antenna! Although from the photo I'm not sure he bought the cheap antennas.
Speaking from experience deploying stratum-1. Not stratum-0, but the same GPS concerns.
I've looked at doing this a few times. I don't have references handy but there are cheaper atomic oscillators available now, under $1000. Still to expensive for me to justify it but one of these years I'll find one cheap.
I have an NTP clock that uses GPS PPS and has a local TXDO. For any reasonable amount of time that you'd be out of GPS (barring nuclear war) a TXDO should be plenty sufficient for any sane time-related needs at a tiny fraction of the cost. Serving as a frequency reference for radio or precision counters or other semi-exotic (for at home) is the only reason I can see to actually have CSAC other than cool factor. Which, fair.
That 'local TXDO' is most likely being governed by some kind of phase lock on the PPS or some other divisor and may well lead to terrible clock jitter in the short term while giving you insane precision over the longer term. Allan deviation plots made by a device clocked by a standard that is better than the one that you are checking is the only way to be sure how good (or bad) things really are.
I have a couple of GPSDOs here and it is fun to play them against each other, the differences in the short term can be substantial, but over anything more than a few days they are extremely accurate.
I'm aware, and yeah GPS is disciplining the TXCO, but I'm saying it's stable enough for non specialty uses outside of stuff like serving as a frequency reference. If you just want a clock source for NTP without the internet then all of this is already overkill.
You're also not likely to be out of GPS for particularly long stretches of time.
Ensembling clocks allows you to spot 'the odd one out' in case any one of them is having trouble so in that sense it would allow you to improve precision overall (or rather: availability of precision).
No mention of the cost of the CSAC GPSDO (only that it's "not cheap").
Too bad you couldn't hack the Americium module from a smoke detector and create a DIY atomic oscillator. Cesium seems to be preferred. (And I know nothing about this sort of thing.)
(EDIT: chatting with an LLM… I realize I had assumed that "atomic clocks" meant radioactive and so suggested Americium because it is easy to obtain. LLM schooled me and suggested "Rubidium oscillator modules" instead since they come up for a few hundred dollars or so on eBay. Still not the DIY approach I had hoped for—I think I am still channelling the old "Amateur Scientist" column from Scientific American from the day.)
Typically high hundreds / low thousands for a used GPSDO with a CSAC.
And Americium is not as useful for a timing reference, as it's not as stable as Rubidium and a lot less safe to handle. Otherwise time nuts would hoard cheap smoke detectors :)
I have 2 Rb GPSDOs and a few more OCXO ones, but had never heard of CSAC modules and thought that the inevitable next step would be an HP 5061A Cs clock or later model.
So now you have me going to eBay in search one but all it turns up are BM25CSAC carburetors! What are the magic keywords to use in my search?
Search for "rubidium clock" or "gpsdo rubidium" and you'll start whittling it down more. It's much harder to find just the CSAC part, unless you're building clocks commercially.
The clocks need precise oscillations to measure time: the mechanical clocks used pendulums and springs; the maser clocks used precise microwave cavities; the atomic clocks that are dominating today use oscillations of electrons in selected atoms.
Americium is not a good atomic clock material---it doesn't have superior electronic transitions, and the nuclear transitions causing its radioactivity would get in the way.
Nuclear oscillations could also be used: there is a proposal to use a low-energy nuclear oscillation in Thorium; it would be more stable than electronic oscillations:
https://en.wikipedia.org/wiki/Nuclear_clock
The distinction of what can or cannot be called 'radioactive' is somehow artificial: masers, atoms and nuclei all emit radiation, so they all are technically 'radioactive'. Conventionally, 'radioactive' radiation requires energies that cause ionization of common materials, usually quoted as above 10eV. The Thorium nuclear transition is actually below that, so technically it is not radioactive---but I'd still not want to sit next to such clock without some shielding, because even UV radiation with energies above 3eV is known to damage living tissue.
He seems to be using a GPS-2700[0], which has a price tag of about $5500 / €4700. I reckon you can find a better price if you get very lucky on the second-hand market, though.
It strikes me as strange that the article links to [1] which appears to be the same board, absent the "Viavi" logo on the main RF can, as the Microchip product you linked. I couldn't tell with a brief look if the Viavi product is offering something like software, configuration, tuning, etc. on top of GPS-2700 product.
The photo of the device on the article says "Jackson Labs" which seems to have been the previous name of "Viavi Solutions" and a review video [2] mentioned using Symmetricom atomic clock modules, which was acquired first by Microsemi (2013) and subsequently Microchip (2018)[3].
There are some subtle differences. The Jackson Labs and Microchip boards both have a diagonal "swoosh" and a "do not touch" icon on the metal clock casing, a u-blox branded GPS receiver, and partially-filled mounting holes. The Viavi board has a blank clock casing, unbranded GPS receiver, and fully drilled-out mounting holes. But yes, all three are using a virtually-identical PCB.
Judging by the misaligned capacitors(?) on the Viavi board, it is almost like the Viavi one is an early prototype, with the Jackson Labs one being an early production version and the Microchip one being the current production version. I have no idea how that would work out acquisition timeline wise.
But yeah, hardware companies are rather acquisition happy. When designing hardware it is very common to come across datasheets with an "X is now known as Y" cover page stapled onto it. Heck, every once in a while you'll even come across a datasheet which is obviously scanned-in, for a brand which hasn't existed in three decades - and the chip will still be in production!
I don't believe it's necessary to have multiple GPS antennas (one per device), unless signal path redundancy is required. A good GPS distribution box like from Time Machines or GPS Source can split the antenna signal to many devices without an issue.
A signal distribution box used from eBay is a lot cheaper than a good outdoor GPS antenna!
Though if you have enough cable and enough antennas already, no harm in having a little array like in OP.
>I don't believe it's necessary to have multiple GPS antennas (one per device), unless signal path redundancy is required
Don't those splitters typically have an amplifier, so become a single point of failure?
If you're going through the trouble to have multiple time servers on your network, you probably want to make sure that an amplifier failure doesn't take them all down at the same time.
The antenna has an amplifier, usually with more than enough gain to support downstream splitting. But then sure the antenna is a SPOF.
Also the splitter will pass through the dc-bias on the first port to power the antenna, -- so whatever is hooked there is another spof.
Depending on your application a GPSDO may well be the entirely wrong solution depending on lots of implementation details.
GPS antennas typically need power. GPS receivers expect to be the sole power supplier. Simple, cheap signal distribution won't handle this. There are splitters that can allow one downstream port to supply power, but that fails when that single receiver fails. There are other splitters that are active-active. I guess you're claiming that the boxes from Time Machines are such active-active devices.
Those are not cheaper than an antenna! Although from the photo I'm not sure he bought the cheap antennas.
Speaking from experience deploying stratum-1. Not stratum-0, but the same GPS concerns.
(2017)
I've looked at doing this a few times. I don't have references handy but there are cheaper atomic oscillators available now, under $1000. Still to expensive for me to justify it but one of these years I'll find one cheap.
I have an NTP clock that uses GPS PPS and has a local TXDO. For any reasonable amount of time that you'd be out of GPS (barring nuclear war) a TXDO should be plenty sufficient for any sane time-related needs at a tiny fraction of the cost. Serving as a frequency reference for radio or precision counters or other semi-exotic (for at home) is the only reason I can see to actually have CSAC other than cool factor. Which, fair.
That 'local TXDO' is most likely being governed by some kind of phase lock on the PPS or some other divisor and may well lead to terrible clock jitter in the short term while giving you insane precision over the longer term. Allan deviation plots made by a device clocked by a standard that is better than the one that you are checking is the only way to be sure how good (or bad) things really are.
I have a couple of GPSDOs here and it is fun to play them against each other, the differences in the short term can be substantial, but over anything more than a few days they are extremely accurate.
I'm aware, and yeah GPS is disciplining the TXCO, but I'm saying it's stable enough for non specialty uses outside of stuff like serving as a frequency reference. If you just want a clock source for NTP without the internet then all of this is already overkill.
You're also not likely to be out of GPS for particularly long stretches of time.
Go hunt up a Rubidinium standard on Ebay, $150 or thereabouts and you'll have a pretty good standard.
If you use the GPIO-based PPS on RT Linux, the kernel will soon get a nice improvement in jitter: https://lore.kernel.org/all/A0383FF7-267A-4D4C-9B74-F80B9908...
I've been testing it and it is noticeably better, but I haven't yet had time to set up an A/B test to get real numbers.
I wonder how, if at all, you can improve precision with 4 stratum-1 clocks like he author has.
Ensembling clocks allows you to spot 'the odd one out' in case any one of them is having trouble so in that sense it would allow you to improve precision overall (or rather: availability of precision).
No mention of the cost of the CSAC GPSDO (only that it's "not cheap").
Too bad you couldn't hack the Americium module from a smoke detector and create a DIY atomic oscillator. Cesium seems to be preferred. (And I know nothing about this sort of thing.)
(EDIT: chatting with an LLM… I realize I had assumed that "atomic clocks" meant radioactive and so suggested Americium because it is easy to obtain. LLM schooled me and suggested "Rubidium oscillator modules" instead since they come up for a few hundred dollars or so on eBay. Still not the DIY approach I had hoped for—I think I am still channelling the old "Amateur Scientist" column from Scientific American from the day.)
Typically high hundreds / low thousands for a used GPSDO with a CSAC.
And Americium is not as useful for a timing reference, as it's not as stable as Rubidium and a lot less safe to handle. Otherwise time nuts would hoard cheap smoke detectors :)
I have 2 Rb GPSDOs and a few more OCXO ones, but had never heard of CSAC modules and thought that the inevitable next step would be an HP 5061A Cs clock or later model.
So now you have me going to eBay in search one but all it turns up are BM25CSAC carburetors! What are the magic keywords to use in my search?
Search for "rubidium clock" or "gpsdo rubidium" and you'll start whittling it down more. It's much harder to find just the CSAC part, unless you're building clocks commercially.
Americium is also chemically different and may not be as practical as cesium and rubidium to ionize and trap.
Words of wisdom (from a time nut).
Spinthariscope https://xkcd.com/2568/
The clocks need precise oscillations to measure time: the mechanical clocks used pendulums and springs; the maser clocks used precise microwave cavities; the atomic clocks that are dominating today use oscillations of electrons in selected atoms.
Americium is not a good atomic clock material---it doesn't have superior electronic transitions, and the nuclear transitions causing its radioactivity would get in the way.
Nuclear oscillations could also be used: there is a proposal to use a low-energy nuclear oscillation in Thorium; it would be more stable than electronic oscillations: https://en.wikipedia.org/wiki/Nuclear_clock
The distinction of what can or cannot be called 'radioactive' is somehow artificial: masers, atoms and nuclei all emit radiation, so they all are technically 'radioactive'. Conventionally, 'radioactive' radiation requires energies that cause ionization of common materials, usually quoted as above 10eV. The Thorium nuclear transition is actually below that, so technically it is not radioactive---but I'd still not want to sit next to such clock without some shielding, because even UV radiation with energies above 3eV is known to damage living tissue.
He seems to be using a GPS-2700[0], which has a price tag of about $5500 / €4700. I reckon you can find a better price if you get very lucky on the second-hand market, though.
[0]: https://www.microchip.com/en-us/product/GPS-2700
It strikes me as strange that the article links to [1] which appears to be the same board, absent the "Viavi" logo on the main RF can, as the Microchip product you linked. I couldn't tell with a brief look if the Viavi product is offering something like software, configuration, tuning, etc. on top of GPS-2700 product.
The photo of the device on the article says "Jackson Labs" which seems to have been the previous name of "Viavi Solutions" and a review video [2] mentioned using Symmetricom atomic clock modules, which was acquired first by Microsemi (2013) and subsequently Microchip (2018)[3].
[1] https://www.viavisolutions.com/en-us/products/chip-scale-ato...
[2] https://www.youtube.com/watch?v=CogN630jUSs
[3] https://en.wikipedia.org/wiki/Symmetricom
There are some subtle differences. The Jackson Labs and Microchip boards both have a diagonal "swoosh" and a "do not touch" icon on the metal clock casing, a u-blox branded GPS receiver, and partially-filled mounting holes. The Viavi board has a blank clock casing, unbranded GPS receiver, and fully drilled-out mounting holes. But yes, all three are using a virtually-identical PCB.
Judging by the misaligned capacitors(?) on the Viavi board, it is almost like the Viavi one is an early prototype, with the Jackson Labs one being an early production version and the Microchip one being the current production version. I have no idea how that would work out acquisition timeline wise.
But yeah, hardware companies are rather acquisition happy. When designing hardware it is very common to come across datasheets with an "X is now known as Y" cover page stapled onto it. Heck, every once in a while you'll even come across a datasheet which is obviously scanned-in, for a brand which hasn't existed in three decades - and the chip will still be in production!