UltraTrac on 945 VERY, incredibly inaccurate (cycling) for activities?

pgrey said:

Why wouldn't UltraTrac be "for" an activity that isn't "continuous", necessarily?  Is this actually the documented purpose?

I have no idea. You could argue the same for any of the GPS settings. However, experience has shown that trying to accumulate multiple days of activity into on on any Garmin has resulted in a mess.

But I am not convinced that there is a bug as much as using it for a purpose for which it most likely was not intended.

Best people to talk to clear this up for you and the OP is Garmin. 

I had to try out ultratrac on my Fenix 5, well aware of the fact that it may have a different algorithm, as well as has a different GPS-chip and antenna construction compared to the 945.

I used smart recording, and logged my short daily commute on bike to the office. The distance usually records to some 7.4 km, using GPS and 1 second recording. I did get 6.86 km using ultratrac/smart. It is a short distance with several stops for crossings/red lights, two short tunnels and a couple of fairly sharp turns as well as some longer stretches without turns.

Keeping an eye on the watch (that I could not do all the time, obviously) I recognized that GPS was connected fully during the first two minutes, and after that 20 seconds per full minute with a bit of a variance. I assume that it uses an algorithm taking measured speed during GPS lock and possibly movements of the watch into consideration. It can after all detect cycling as an activity using Move IQ.

During the short commute, the distance jumped up every time GPS was acquired (expected), and was at all times some 10% within actual distance (although always short). Looking at the route, it is obvious that it cuts corners between GPS lock (also expected).

The displayed speed was given as actual during the 20 seconds of GPS lock, after which it froze on the most recent speed for some seconds followed by no speed at all “- -“ and no change of distance displayed.

If there is a learning algorithm being used to increase the precision of ultratrac, it has had plenty of time to learn my patterns since I have had the F5 since March 2017, and used it daily for plenty of different activities.

Of course, this is just one try, so that must be taken into consideration but I wanted to share my experience of biking and ultratrac on a route that I have tracked many, many times. And of course NOT with an 945, but it may be of interest anyway.

Thanks for sharing, always interesting...

Would you mind sharing the garmin connect link? (Making it public?) Wondering what it looks like...

Well, I mentioned in my post that there were some 'buts' right? And it's not as if I happen to have a 40 hr activity to do to just test ultratac 'properly'.

And from the 945-manual:

The UltraTrac feature is a GPS setting that records track points and sensor data less frequently. Enabling the UltraTrac feature increases battery life but decreases the quality of recorded activities. You should use the UltraTrac feature for activities that demand longer battery life and for which frequent sensor data updates are less important

So sure, a 30 minute training is no valid option (still, shouldn't matter for battery usage) but frmo this you could conclude biking should be perfectly fine...

Interesting, but I would have expected it to be even closer to the actual distance based on how clean your track is.  I don't see enough variance from the road to account for a 'connect the dot' service to miss a couple of miles.  Regardless, I think this demonstrates how you believe it should work versus how it is designed to work.

I really haven't been able to validate that the 935 behaves any differently when cycling.  Certainly the 920 followed the current behavior pattern and so I find it odd that the 935 would be any different.  I have a friend with a 935 so I'll try and pry it from them for a day or two and so some testing.  I'd be really surprised if the results indicated a design change.

I'm still in the camp that it works remarkably well in certain applications.  Cycling isn't one of those since you aren't able to maximize stride length.  I can't address the battery drain on the 945 but I get acceptable battery consumption using the Fenix 3.  I had quite a bit of documented results on the old forums but they've been removed so there's that.

Considering that it seems to work as intended I don't know if the designers would respond to any bug incidents.  That's a limitation of WDR but it's by design and works well if you appreciate the benefits/limitations of that technology.

On a side note, are you using Smart or 1 Second data recording?  Maybe we can help Flip diagnose if Smart is recording data points intelligently.

Sorry, prefer privacy in this case. I can only offer my explanation of the track: it is quite good, with the exception of sharp corners. The track is pretty much following my actual route. I tried ultratrac on the return trip too, giving 6.84 km, almost identical to the first trip. And it should be, since it it the same tour, just the other side of the road. Still a bit short compared to 7.4 km, though. But consistent. 

No worries and thanks for the extra info...

In my bike case (few posts above) the track was pretty ok but it missed a lot of the distance, there were also many gaps in the speed graph so that was what I mainly was wondering about... but 6.9 out of 7.5 seems fair indeed for ultratac

• Zero speed between points is obviously a bug. Otherwise it's useless for tracking events. You're trying to excuse the inexcusable. It's a less frequent tracking mode that compromises pin point accuracy for extended battery life. No ifs or buts. If it's NOT a bug then it's being missold and many of us are going to be very annoyed. 

Yes. I did run for 18.5km with a race GPS on my bad, watch reported 99.6km, real distance was 85.6km. Ultratrac is really a fail. On my 10km city race it reported 12.34km, with stupid spikes to 2 blocks away for no reason https://www.strava.com/activities/2668493685