In reply to Wibble Wibble:
> I used a GPS in a white-out in Iceland on a glacial plateau. The direction arrow fluctated wildly (like 180 degree direction changes) whilst attempting to move towards a waypoint some 2km away.
Exactly what I found trying to navigate off Stuc a'Chroin.
> Any explanation for this?
Yes. The course (ie current bearing) indication basically works by comparing where the GPS thinks it is now to where it thought it was last time it looked. Depending on how often it checks, and how fast or slow you are moving, the positional error can easily exceed the distance you've actually covered, and in a random direction.
For example, andy says that most GPS update their course indication every second (I haven't managed to find any specific information online to confirm this). At Naismith rule walking speed of 5kmph, you cover 1.39m in one second. Garmin quote the positional accuracy of their receivers as being 15m. Even if it were 10x better (which even with WAAS it isn't - Garmin says 5x better on average), that would still be enough to throw the apparent course out by up to 180 degrees (ie if the positional error were in the opposite direction to the one you'd actually travelled). Which is what we've both seen.
They could use some averaging methods to improve the accuracy of the indicated course but I'm not sure that they do. I certainly can't find anything about it online. Basically, the manufacturers don't seem too keen to go in to detail on this point, probably because (wait for it)
it's not a primary function of the device. GPS is a
positioning system (it's in the name!) - anything else it offers is a derivative from that.
By the way, this is also the reason that you will find that the trip distance reported when walking is often way out. The trip meter does use an averaging mechanism to smooth out the errors, otherwise the trip distance would tend to be too large because the positional errors can make it look like you're zig-zagging all over the place. That works OK at, say, driving speeds but again it's not smart enough to cope with the relatively small changes of position between samples at walking speed. I can derive significantly more accurate trip distances from my eTrex by applying Pythagoras between the trackpoints on my tracklog (in Excel, after I've got home and downloaded it to my Mac!) That's because I've set my tracklog to sample every thirty seconds, which means that at walking speed the positional error becomes less significant.
This is the sort of thing I mean by "understanding the limitations of your equipment".