From my understanding of GNSS (I did some research on it in school, about a month full-time), that only works if you have a clock synchronized to a few nanoseconds and of course up to date satellite information. This is very plausible if you do it once a second, because in 1 second your quartz doesn't drift that much or you can count cpu clock cycles perhaps, and the ephemerides don't change that often.

But once you move indoors and lose signal regularly and have a much higher noise to signal ratio, I very much doubt this can still work. I don't know about you, but I don't work in the garden, so most of the day that GNSS unit would still be listening and calculating (as I understand, the code division multiplexing means you kind of brute force the various codes until one produces a signal instead of noise) and might not even get a fix at all (my phone takes a few minutes next to a window when there is concrete on all sides except for that narrow view of the sky, and this 2018 phone is already orders of magnitude better than my previous Galaxy Note 2 was). I'm not saying it's impossible, but I would highly doubt that it consistently gets a fix in 2ms, or even 200ms, in the place where you spend most of your time.

Of course, it's mostly useless indoors anyway, but so you'd have to manually toggle it every time you go in- or outside, or somehow configure a home location where it turns off (would work great during COVID) if it has such a feature, though I wouldn't know how it would figure out when you've left home in that case. (WiFi comes to mind as an obvious solution, but then how do you explain to the user that it won't track your GPS while that WiFi is in range? It just seems like they'd drop the auto-off feature at this point.)

I would assume a common use for GPS in a watch is workout tracking, so it gets manually started and stopped when they go for a run. it'd need more time at the start to get a clean fix, but in good conditions could probably get by with short blips.

Ah, I guess it shows that I'm no fitness tracker user! I didn't realize you would already interact with the device to indicate you're starting an activity. Yep, in that case this is very plausible, and on startup it probably takes only a few seconds of continuous searching if you have as good a receiver as the 2ms claim indicates.

It is possible to do the GPS processing after the fact. You turn on the rx and listen for a few hundred milliseconds, then turn off. This is very low power. Later you push these to a server (a smartphone or the cloud) which computes the position. This is called snapshot processing.

So if you don't need the position in real time for navigation it is quite possible to track location over a week with a tiny battery. This could be time triggered and/or based on MEMs counters (which are also power friendly). Or you could calculate locally if the user explicitly asks.

https://ieeexplore.ieee.org/document/5708033 http://publica.fraunhofer.de/documents/N-464241.html