GPS bugs on Suunto9

Fenr1r

@Lukasz-Berdowski Right. So. Going by your summary of repeated observations and perhaps as many as four Google Streetview shots …

Clearly each watch is happier with tree-partheid. From the tracks shown, both devices lose it in areas of mixed deciduous (light green in sat pic. Ash? Birch?) and evergreen (dark green: a fir of some sort?). Without the mix (NE), or with the species segregated (SE), they do better/fine. The TomTom is obviously less disconcerted by arboreal mixing. The Suunto freaks out when almost wholly surrounded by evergreens and takes some time to recover.

Different tree types’ interference may be a thing. If the S9B has a clever compensatory algorithm for adjusting to/anticipating GNSS signal quality, the varying mixture might cause more knock-on problems than the clear division.

(Or it might be track width and overhead foliage density. Or something else entirely.)

Łukasz Szmigiel · S9PP 2.39.20

@Fenr1r trees are always problematic in terms of GNSS accuracy but I’ve ~~never~~ (well once the watch got completely bananas but it was with older FW) observed such problems in my S5. If there is a problem it typically manifests itself with offset which is pretty much constant (ie. 10 - 20 meters off) untill I change direction or signal improves (opening in the tree line, etc.) and this is ok.

But the image attached by @Lukasz-Berdowski indicates a rather big error in first 2 km of run.

@Lukasz-Berdowski have you tried soft reset, fresh AGPS sync, 15 minutes to unpack AGPS and then ~60 seconds on open area after acquiring fix?

Fenr1r

@Łukasz-Szmigiel Could the S5 be dumber? Presumably (a tiny bit) more like the TomTom? The S9 suffers from its own post-processing cleverness?

Trees - yup. In the image, the wild errors really correlate with the tracks’ (in-)visibility to Google’s aerial photography (whenever in 2020 that happened): i.e., sufficient overhead coverage. That might be enough.

A bit boring and perhaps not applicable to that SE stretch from 4km, so I added the tree type factor. The SE track is partially visible but that’s still sort of a canyon effect so sat geometry is trickier for a regularly acquired series of good fixes.

Again: might be utterly irrelevant.

Łukasz Szmigiel · S9PP 2.39.20

@Fenr1r S5 / S9 use the same GNSS chip don’t they? S5 is supposed to be better in terms of signal reception due to external antenna but it’s hard to judge without debug data (raw data from sensor).

Also, could it be related to FusedTrack? Or is it deactivated in performance mode?

Fenr1r

@Łukasz-Szmigiel said in GPS bugs on Suunto9:

Also, could it be related to FusedTrack? Or is it deactivated in performance mode?

Just what I was wondering re prediction/compensation. S5/S9 - same firmware using that same chip in the same way, whether or not FusedTrack active?

Łukasz Szmigiel · S9PP 2.39.20

@Fenr1r there’s no FusedTrack in S5. The question is - is it hardcoded that FusedTrack is disabled in performance mode or does it activate automatically if GNSS signal is poor?

Fenr1r

@Łukasz-Szmigiel And it’s a good question. And if the answer is Yes (=> No to the second part): is there anything else that S9 might be doing that might be particularly affected by type of canopy? Anything that your S5/L-B’s TomTom Spark wouldn’t?

Lukasz Berdowski

Wow guys, thank you so much for trying to help me understand the problem To be honest, I don’t know if the fusedtrack is active or disabled while being on the Performance mode. If yes, that would certainly be a good explanation.

@Łukasz-Szmigiel I havent done the reset, just the final 2 steps. By the way, I assume you are in Poland too, what GPS mode do you find to be the most accurate?

There is one more thing - sometimes I feel a very small, almost unnoticable watch vibration (lasting maybe a 0,5 sec) during runs but there is no message on the watch face. I was reading somewhere however, that that is the barometer

Łukasz Szmigiel · S9PP 2.39.20

@Lukasz-Berdowski I usually stick with galileo. In dense forests and mountains glonass is also handy. Beidou seems to give the best response (quickest to get back on track) after track accuracy degenerates but also gives me strange offsets. Also I don’t find it so good in woodlands. Generally with latest firmware, in the open, plain gps is more than enough. But the error gets pretty significant as soon as I enter wooded area so I usually use galileo - the error is there but it’s more like an offset than a zigzag.

I’ve been also setting a description of which GNSS I use in my workouts to have a better look on it and more data.

Dimitrios Kanellopoulos

Fused track is 100% off on best mode.

We had that on before the watch got into production and the results were bad because in many cases the compass is not calibrated.

I have seen your tracks. They are not good. I also wonder why.

Łukasz Szmigiel · S9PP 2.39.20

There is definitely some magic going on with how watch interprets raw signals from GPS sensor which may lead to issues in rare conditions. For example, when stationary, S5/S9 doesn’t draw motion - there are no weird shapes being drawn on map (which isn’t the case with Garmins, for example, as they’re constantly drawing).

So there must be some filtering which may be failing in certain conditions (as in this post). It’s hard to guess without knowing the actual methods Suunto is using. By analyzing my own tracks I’m guessing FusedSpeed is influencing the recording (if pace is prioritized over GPS accuracy) but it manifests itself with track offsets (that are constant until change of directions). But that’s my guess anyway.

Also, in my experience, the longer the run, the better the path being drawn - perhaps there’s some room to optimize this behavior?

Dimitrios Kanellopoulos

@Łukasz-Szmigiel The “not-draw” when static is Sony magic.

Łukasz Szmigiel · S9PP 2.39.20

@Dimitrios-Kanellopoulos cool. I thought it’s Suunto’s algo. So the chip doesn’t even provide raw data? It’s always processed?

Dimitrios Kanellopoulos

@Łukasz-Szmigiel I am not sure about that, but I can say that the chip indeed does a lot on it’s own.

Lukasz Berdowski

Hey Guys, I just did a new run, results are much better, while still not perfect, I can live with it

I did a soft reset, updated gps via app, waited 15 minutes, then acquired the gps signal and waited additional 3-5min.
The run was tracked with GPS & Galileo.

trasa 3.JPG

sartoric

@Lukasz-Berdowski
this one looks good.
Consider that the “worst” parts are really into the woods

Lukasz Berdowski

@sartoric thank you, I agree. I will keep testing, nad hopefully the above mentioned process will not dissapoint

Also, I felt very slight singular vibrations on the watch more or less where the biggest inconsistencies are. Maybe the GPS is loosing conncetion and recconecting or something? It happens every time on this track. But there is no message being displyed

sartoric

@Lukasz-Berdowski said in GPS bugs on Suunto9:

I felt very slight singular vibrations on the watch

Are you navigating a route ? Could be the “off route” message

there is also the chance that the path slightly changed compared to the satellite photo as both tomtom and s9 track it on the right (the upper left section)

Łukasz Szmigiel · S9PP 2.39.20

@Lukasz-Berdowski this looks pretty ok. You can also try glonass, its satellites have different inclination than GPS and Galileo and there are (usually) more sats available than Galileo so it may also be helpful in dense forests.

I also wonder how will the recorded path look like if you reverse directions?

Dimitrios Kanellopoulos

@Lukasz-Berdowski what hand do you wear the watch ?

And what’s the direction of the route I see ? Clock or counter clockwise ?