Forerunner 935 GPS Accuracy/Performance

Also need to be clear what we mean by "accuracy" in this context.

Those articles largely talk about accuracy of an individual point. Rarely is this the major issue in a sports watch when we typically talk about the accuracy in distance accumulation although of course it is "nice" if the GPS track puts you on the right side of the road! This may not always be the case however given Garmin themselves say "Garmin GPS receivers are typically accurate to within 10 meters."

On a track that means if you ran in lane 1 but the GPS track showed you in lane 3 it is not unexpected.

And then further "typically" suggests that sometimes some points along your way may be a bit over 10 metres out.

Individual points in the track aren't completely independent, anyway, and errors in distance don't accumulate as fast as you might think.

One of the error sources is atmospheric ionization, which doesn't change very quickly, so is similar in effect in neighbouring points and, actually, for quite some distance.

GPS receivers don't calculate each point completely independently; they use a process called Kalman filtering, which estimates the receiver speed from the Doppler shift of the satellite signals and calculates the next position as a combination of the previous position, previous speed, and the GPS solution for the current position. Generally it'll use multiple points for this. It'll reduce the amount of jitter in location from point to point. Both this, and the following point, are more effective at higher speeds, which is one reason cycling works better than running, which works better than walking. (Another reason is that it's easier to reject multipath signals at higher speeds, because the Doppler effect makes it easier to identify them).

As far as cumulative errors go, provided that the jitter is fairly small compared with the distance from one point to the next, errors parallel to your direction don't accumulate at all, though they do contribute to noise in instance pace, and errors perpendicular to it have a fairly small effect which you can estimate from Pythagoras' equation.

TLDR, just because one individual point might be 10m off does not mean that every point is going to be 10m off in a random direction.

And in the same vein, just because one track is bad (or good) on one device doesn't mean that all tracks on every device are going to be bad (or good). There're too many variables involved to expect better than stated performance every time. Yes be happy when it happens, but there are no grounds for complaints when it doesn't, provided of course it is still within specification.

Unless of course there is something horribly wrong with the device itself.

My point remains. Many of the people who complain about poor GPS accuracy have unrealistic expectations. Many of the people who complain about poor GPS accuracy are more concerned with the displayed track than the distance measured. I too can zoom in far enough to see the hairs on a gnat's rear end and see my tracks go wonky. But, since I run mostly on trail anyway I don't care. The distance I get on the routes I run is repeatable. When I enter events, the distance I end up with is within cooee of the stated distance. And when I do run on road, the same thing applies.

And don't get me going on the instant pace discussion!

Wow, thank you McB. This probably explains why the very satisfying Virtual Training Partner on the bike turns into a random training partner on the trail. I'm wondering if a virtual power partner might side step these issues altogether.

They are not ALL complaining about GPS tracks.

I never said they were.

"The distance I get on the routes I run is repeatable. When I enter events, the distance I end up with is within cooee of the stated distance". So why shouldn't others expect the same?

No reason at all why they shouldn't. But there should not be an expectation that under all conditions, they will get identical results every time, or better than the specifications all of the time. I'd be upset if I continually got poor results (>5%) all of the time. But as it happens only occasionally I don't get upset and post here every time it does happen.

I expect less then 1% error

which, based on Garmin's stated specifications is unrealistic, especially if you expect that to occur every time you run under every condition, on every different type of surface, in every different environment.

And that has been my point all along

And in the same vein, just because one track is bad (or good) on one device doesn't mean that all tracks on every device are going to be bad (or good). There're too many variables involved to expect better than stated performance every time. Yes be happy when it happens, but there are no grounds for complaints when it doesn't, provided of course it is still within specification.

Unless of course there is something horribly wrong with the device itself.

My point remains. Many of the people who complain about poor GPS accuracy have unrealistic expectations. Many of the people who complain about poor GPS accuracy are more concerned with the displayed track than the distance measured. I too can zoom in far enough to see the hairs on a gnat's rear end and see my tracks go wonky. But, since I run mostly on trail anyway I don't care. The distance I get on the routes I run is repeatable. When I enter events, the distance I end up with is within cooee of the stated distance. And when I do run on road, the same thing applies.

And don't get me going on the instant pace discussion!

Well, another point remains as well. People have come to expect a certain performance based on previous experiences...

Purely from a gps-spec point of view you might not expect as much, but the companies are pretty damn clever and have proved they can work around a lot of these issues using other sensors, algorithms, etc. When my previous watches (tomtom, garmin 235, 735xt) performed on a certain level I kind of expect the newest garmin flagship (fenix 5) to be able to perform on that level as well. Is that unrealistic?

Most complaints are not about the path not following them to the meter, they are more general. My Fenix5 was pretty bad gps-wise, compared to my other experiences. Perhaps it was within official GPS spec, but it was not within what I have come to expect from a product like this.

Depending on how you use it this might not be an issue, but for some it is. I ran a half marathon with both the 735xt and the fenix5, at a certain points the fenix5 was pretty off at km-marks. (The 735xt was the correct one in this). In the end the difference between the 2 was 200m, only 1% so pretty damn close but still. The fenix5 was ok, but had much more variation continuously... adding up in the end. Man, if 3-5% is acceptable that would mean an entire 1km on the hm, can't imagine anyone actually finding that acceptable.

The instant pace is simply a consequence of this. I don't use instant pace, but even lap pace takes much longer to settle on the F5. The 935 gives a much better indication much faster. It's not perfect but it's actually usable.

And then you can say 'if you want accurate pace buy a footpad'. Sure, and if you want accurate hr buy a strap, etc. I buy this device because I want to use as little extras as possible. I hate a hr-strap, I don't want to be busy with footpods (yet). I use a watch that has it all so I don't have to, as a trade-off I don't expect the best but I do expect decent and usable.

Where the other watches delivered on this (knowing the limitations) the fenix5 didn't. So I complained about that... and returned it. Now the 935 DOES perform as expected meaning these expectations aren't too unrealistic right?

Anyway, people can be happy and unhappy. All in their own rights, but to simply write complaints off as 'unrealistic expectations' is nonsense... Just as writing off happy reports as 'well, bias'.

I have owned a number of GPS watches over the years, different generations of Garmin products and also products from other manufacturers. The results I am seeing from my 935 are as good or better than any other GPS watch I have used in the past (but I have never been vain enough to choose form over function and buy a fenix).

There might be an element of luck involved here. The specification for a GPS receiver chip will state a certain minimum sensitivity but there is always some variation in that. You could be the lucky one who has a relatively sensitive chip, or the unlucky one who has a chip that only just made the minimum spec. On a clear day in the open both will perform fine, on a rainy day under tree cover this reduction in sensitivity could make all the difference. Maybe I have a lucky 935 and my user experience is a lot more positive than yours. Unless we can test them next to each other we simply don't know as the end users how our 935s compare.

This is a link to a run I did last Sunday. Everything was in the 935's favour. Clear sky, rural setting, dry, not a lot of overhanging vegetation. It was a certified 10k road race, my 935 said 9.99k. Zooming in on the track I can almost tell where I drifted to the right side of the road (in UK road races you are meant to stay on the left) and cut a few meters in a couple of corners. Even the pace graph looks usable (no foot pod on this run). So now I know what my 935 is capable of at its best. But I do not expect the same results every time. There are simply too many factors I cannot control for that to be realistic.

High Easter 10k

@CloggyAbroad: Not sure if that was a reply to me? I am very happy with the 935 myself as well, performs great. My gripes were with the fenix5 (indeed form over function, but at that time I didn't know the functions would be impacted as much). But glad to hear you are a happy 935 owner as well ;)

@FlipStone: not a reply to you as such, just some general observations. But I do believe people have unrealistic expectations when it comes to GPS watches in general