Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

bobval

macrumors newbie
Original poster
Apr 2, 2015
24
9
Maybe someone measured the accuracy of the Apple Watch without the Internet. I'm interested in the error for 7 days.
 
I've seen a test but I'm not really sure that it was totally genuinely offline. The test description wasn't super clear.

It's effectively a quartz watch, but with an oven for the crystal. It should be more accurate than your average Casio.


Quartz watches accuracy - how accurate is quartz watch?

A non-certified modern quartz watch has an accuracy of about 99.9998%. A certified one is 99.9999% accurate.

To put this into perspective, a non-certified quartz watch will lose only around a second per day while a certified one generally won’t lose any. A vintage mechanical watch can lose up to 60 seconds a day, and a modern one between 3-5 seconds.

Over the course of a month, the lost seconds can pile up and become significant.


My guess is less than a second a day of drift.
 
Technically the watch shouldn't need the internet to synchronize the clock, as a time reference accurate to about 30 nanoseconds is present in the signals from the GPS satellites. Pretty much every portable car GPS I've ever used had the ability to derive and display date and time from them, though I'm pretty sure the watch does not.
 
  • Like
Reactions: BlackBun
Technically the watch shouldn't need the internet to synchronize the clock, as a time reference accurate to about 30 nanoseconds is present in the signals from the GPS satellites. Pretty much every portable car GPS I've ever used had the ability to derive and display date and time from them, though I'm pretty sure the watch does not.

I believe it does as a secondary source, I could be wrong though.

I was really looking for an experiment where the watch was stored in a faraday cage for a period of time but I wasn't able to find anything.
 
  • Like
Reactions: skeptech
I believe it does as a secondary source, I could be wrong though.

I was really looking for an experiment where the watch was stored in a faraday cage for a period of time but I wasn't able to find anything.
As an academic, that means you’ve stumbled upon a novel experiment to do 😉
 
  • Haha
Reactions: Howard2k
I ran a simple experiment with an Apple Watch Series 2 (so, sample of one) in 2019 to characterize the internal quartz chronometer in Apple Watch in airplane mode, inside a house which blocks (the somewhat weak) GPS signals. Result: the watch (sitting on its charger 24x7) gained about 1 second in 30 days, and gained less than 2 seconds in 60 days, which puts it within the category of "high accuracy quartz" (HAQ) watches: no more than ±10 seconds drift per year.

Apple Watch has a temperature-compensated quartz oscillator (TCXO), rather than an oven-controlled quartz oscillator (OCXO) because the former likely eats less of the energy (battery) budget of the watch, and the quartz crystal's frequency can be continuously measured (when the watch is online) against a very wide range of authoritative UTC sources, and mathematically compensated in the frequency counter:
  • GNSS (not just GPS),
  • LTE network time (for cellular models),
  • Network Time Protocol (NTP) over WiFi from the Internet,
  • Bluetooth time transfer against the paired iPhone (which has the same or similar access to authoritative UTC).
This measurement is exactly what Unix NTP daemons do: measure the computer's local real-time clock (quartz crystal usually) against NTP servers elsewhere, and calculate local clock drift (and thus required adjustment to maintain relatively close syncrhonization to UTC). I'm more than a little familiar with time & frequency synchronization in part because I'm the guy who set up the first NTP server at apple.com in the late 1980s (when it was a single VAX-11/780 and later VAX-8650 running BSD Unix), the forerunner to time.apple.com.

Accurate time (UTC) is now available everywhere on Earth in more ways than ever before: GNSS satellites are broadcasting good time all the time everywhere, so you don't need the cellular "mobile phone" wireless network or the Internet for good time.

One of the few articles written about this subject with input from Apple VP Kevin Lynch in 2015: Here's how Apple synchronized all your Apple Watches [to UTC]
 
Last edited:
I have a new Apple Watch S2 that I rarely use, I turn it on every 3~6 months to check the battery, I turned it on two days ago.The delay was an hour and a half compared to the actual time. One hour may be due to the change in daylight saving time, but the other half hour is not.
 
I ran a simple experiment with an Apple Watch Series 2 (so, sample of one) in 2019 to characterize the internal quartz chronometer in Apple Watch in airplane mode, inside a house which blocks (the somewhat weak) GPS signals. Result: the watch (sitting on its charger 24x7) gained about 1 second in 30 days, and gained less than 2 seconds in 60 days, which puts it within the category of "high accuracy quartz" (HAQ) watches: no more than ±10 seconds drift per year.

Apple Watch has a temperature-compensated quartz oscillator (TCXO), rather than an oven-controlled quartz oscillator (OCXO) because the former likely eats less of the energy (battery) budget of the watch, and the quartz crystal's frequency can be continuously measured (when the watch is online) against a very wide range of authoritative UTC sources, and mathematically compensated in the frequency counter:
  • GNSS (not just GPS),
  • LTE network time (for cellular models),
  • Network Time Protocol (NTP) over WiFi from the Internet,
  • Bluetooth time transfer against the paired iPhone (which has the same or similar access to authoritative UTC).
This measurement is exactly what Unix NTP daemons do: measure the computer's local real-time clock (quartz crystal usually) against NTP servers elsewhere, and calculate local clock drift (and thus required adjustment to maintain relatively close syncrhonization to UTC). I'm more than a little familiar with time & frequency synchronization in part because I'm the guy who set up the first NTP server at apple.com in the late 1980s (when it was a single VAX-11/780 and later VAX-8650 running BSD Unix), the forerunner to time.apple.com.

Accurate time (UTC) is now available everywhere on Earth in more ways than ever before: GNSS satellites are broadcasting good time all the time everywhere, so you don't need the cellular "mobile phone" wireless network or the Internet for good time.

One of the few articles written about this subject with input from Apple VP Kevin Lynch in 2015: Here's how Apple synchronized all your Apple Watches [to UTC]
Superb!
 
  • Like
Reactions: skeptech
I ran a simple experiment with an Apple Watch Series 2 (so, sample of one) in 2019 to characterize the internal quartz chronometer in Apple Watch in airplane mode, inside a house which blocks (the somewhat weak) GPS signals. Result: the watch (sitting on its charger 24x7) gained about 1 second in 30 days, and gained less than 2 seconds in 60 days, which puts it within the category of "high accuracy quartz" (HAQ) watches: no more than ±10 seconds drift per year.

Apple Watch has a temperature-compensated quartz oscillator (TCXO), rather than an oven-controlled quartz oscillator (OCXO) because the former likely eats less of the energy (battery) budget of the watch, and the quartz crystal's frequency can be continuously measured (when the watch is online) against a very wide range of authoritative UTC sources, and mathematically compensated in the frequency counter:
  • GNSS (not just GPS),
  • LTE network time (for cellular models),
  • Network Time Protocol (NTP) over WiFi from the Internet,
  • Bluetooth time transfer against the paired iPhone (which has the same or similar access to authoritative UTC).
This measurement is exactly what Unix NTP daemons do: measure the computer's local real-time clock (quartz crystal usually) against NTP servers elsewhere, and calculate local clock drift (and thus required adjustment to maintain relatively close syncrhonization to UTC). I'm more than a little familiar with time & frequency synchronization in part because I'm the guy who set up the first NTP server at apple.com in the late 1980s (when it was a single VAX-11/780 and later VAX-8650 running BSD Unix), the forerunner to time.apple.com.

Accurate time (UTC) is now available everywhere on Earth in more ways than ever before: GNSS satellites are broadcasting good time all the time everywhere, so you don't need the cellular "mobile phone" wireless network or the Internet for good time.

One of the few articles written about this subject with input from Apple VP Kevin Lynch in 2015: Here's how Apple synchronized all your Apple Watches [to UTC]
Interesting that Apple was using VAXes for its servers in the late 1980s (this was pre-web, so no apple.com then), as we were also in the academic arena. After we set up our first webservers on VAXes in 1994, we ended up using Process.com software for many years to host our webpages.
I only use my iWatch 8 in airplane mode, as I don't want it ever tethered to the internet or an iPhone. I suppose that in airplane mode, it does not access GPS signals even outdoors. But I don't need my watch to be more accurate than a few seconds of time. I just bought my iWatch 8 several days ago, so I'll see in a month or two what its time is showing with respect to UTC signals online.
 
Interesting that Apple was using VAXes for its servers in the late 1980s (this was pre-web, so no apple.com then), as we were also in the academic arena. After we set up our first webservers on VAXes in 1994, we ended up using Process.com software for many years to host our webpages.
I only use my iWatch 8 in airplane mode, as I don't want it ever tethered to the internet or an iPhone. I suppose that in airplane mode, it does not access GPS signals even outdoors. But I don't need my watch to be more accurate than a few seconds of time. I just bought my iWatch 8 several days ago, so I'll see in a month or two what its time is showing with respect to UTC signals online.

Um, "apple.com" is a "domain name" and it was first registered in 1986, before I took over maintenance of it in July 1988. However, "apple.com" was also the name of Apple's primary Internet-connected computer at that time: the aforementioned DEC VAX, which, among other things, handled all incoming & outgoing Internet e-mail. Our first Cisco Internet router on the outside was named "torii.apple.com" which you'd see in the output of traceroute(8) from elsewhere on the Internet.

The very first "www.apple.com" server was a Mac IIfx running MacHTTP which was installed in the Valley Green 3 machine room, just a few meters away from Apple's Cray X-MP/48 supercomputer (or had we upgraded to the Cray Y-MP/2E by then? I forget …). It was preceded by "ftp.apple.com" (for downloading releases of MacOS & other utility software) that initially ran on a Mac IIcx running A/UX 3.0, and it sat on the floor in my cubicle.

We had a wide variety of server computers: SGIs, Suns, ... whatever would work best for the task at hand. The AppleLink/Internet e-mail gateway was first on a Mac SE/30 running A/UX 3.0, but when traffic became too much for it, I moved the gateway onto an SGI-4D/380 running IRIX - portable software running on Unix systems is a good thing.

Some quick research doesn't find a definitive statement about whether Apple Watch "airplane mode" turns off the GNSS receiver section - the primary purpose of airplane mode in any device is to turn off RF transmitters to avoid interfering with onboard avionics, but geolocation using GNSS does not require any transmissions - the satellites broadcast all their data all the time, and all GNSS receivers receive the data from multiple satellites and then perform a triangulation calculation - they don't have to transmit anything at all for that function.

So, a quick experiment with my Apple Watch Ultra: put in airplane mode, move around outside while observing the lat/long coordinates in the Compass complication, does Apple Watch Ultra maintain location (positioning) accuracy?

From observed behavior, yes: the lat/long coordinates changed as I walked about 50 feet from my back deck, through my house, to my driveway. The Wayfinder face in its lat/long chapter ring display mode claimed an "updated now" timestamp for the lat/long data while I was out on the driveway where Apple Watch Ultra had a clear view of the sky (the satellites above my horizon).

Preliminary conclusion: no, Apple Watch Ultra does not turn off the GNSS receiver section in airplane mode, and as a consequence, Apple Watch Ultra will maintain its stated ±50 millisecond synchronization to UTC in that mode, so long as it can receive GNSS broadcasts (GNSS radio frequency bands tend not to penetrate typical building materials, save for windows, which is why GNSS geolocation usually doesn't function indoors).

I'd expect Apple Watch Series 8 to behave similarly. Anyone else can perform this simple experiment too.
 
Last edited:
I only use my iWatch 8 in airplane mode, as I don't want it ever tethered to the internet or an iPhone. I suppose that in airplane mode, it does not access GPS signals even outdoors. But I don't need my watch to be more accurate than a few seconds of time. I just bought my iWatch 8 several days ago, so I'll see in a month or two what its time is showing with respect to UTC signals online.
1) no such thing as an iWatch.
2) why would you buy an Apple Watch to run it in airplane mode at all times? A $20 multifunction Casio would be far more practical and wouldn’t need daily charging.
 
  • Like
Reactions: skeptech and jetsam
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.