NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Timekeeping in the post-WWV/post-HF world
From: Peter Monta
Date: 2018 Sep 27, 23:52 -0700
From: Peter Monta
Date: 2018 Sep 27, 23:52 -0700
Let's see what the bounds might be on desired clock performance for autonomous celestial navigation. Let's say 30 days maximum voyage time, and we wish the clock not to contribute at all to the CN error budget.
That might mean time good to one second. This corresponds to 0.25 arcmin in angle, which I think is not entirely unreasonable for CN observations. Yes, the horizon is not reliable to this accuracy, but several star sights distributed in azimuth will allow the dip to be estimated and eliminated, assuming it's the same in all directions. (The angles are "pseudoangles", something like GNSS pseudoranges with their attendant common bias.)
Ok, given this fairly optimistic CN model, we need less than one second probable clock error over 30 days. Can a quartz wristwatch get there? Barely. The best ones can. Because of aging, the watch must have some recent rate characterization; if you wait more than a year or so, aging will cause its rate to drift too far, by an unknown and poorly predictable amount. If the watch has an ordinary single 32 kHz tuning-fork crystal, like 99.99% of quartz watches out there, then it will also need to be on your wrist the whole time.
A no-hassle clock, then, needs to be better than a wristwatch. I'd say a GPS-disciplined OCXO is a cheap, well-understood, mature technology that's fire-and-forget and needs no care and feeding. If GPS malfunctions or disappears, the device automatically goes into holdover mode, and 30 days later the expected clock error is much, much less than one second. (Wear a wristwatch too, I guess, as an additional backup.)
UT1 is predictable to ~10 ms accuracy 30 days in the future, so that's not a problem at all. Maybe it was slightly pedantic to mention it.
Cheers,
Peter
That might mean time good to one second. This corresponds to 0.25 arcmin in angle, which I think is not entirely unreasonable for CN observations. Yes, the horizon is not reliable to this accuracy, but several star sights distributed in azimuth will allow the dip to be estimated and eliminated, assuming it's the same in all directions. (The angles are "pseudoangles", something like GNSS pseudoranges with their attendant common bias.)
Ok, given this fairly optimistic CN model, we need less than one second probable clock error over 30 days. Can a quartz wristwatch get there? Barely. The best ones can. Because of aging, the watch must have some recent rate characterization; if you wait more than a year or so, aging will cause its rate to drift too far, by an unknown and poorly predictable amount. If the watch has an ordinary single 32 kHz tuning-fork crystal, like 99.99% of quartz watches out there, then it will also need to be on your wrist the whole time.
A no-hassle clock, then, needs to be better than a wristwatch. I'd say a GPS-disciplined OCXO is a cheap, well-understood, mature technology that's fire-and-forget and needs no care and feeding. If GPS malfunctions or disappears, the device automatically goes into holdover mode, and 30 days later the expected clock error is much, much less than one second. (Wear a wristwatch too, I guess, as an additional backup.)
UT1 is predictable to ~10 ms accuracy 30 days in the future, so that's not a problem at all. Maybe it was slightly pedantic to mention it.
Cheers,
Peter
