Feedback for GPS 2.60

I get result very similar to your 935; I don't know why, maybe I'm lucky?:)

Hi,

since one week I'am owner of the FR945. I have pretty goods results with GPS accuracy (GPS only) comparing to FR935. I could say more: the FR945 is more precisely in some cases (some corners). Terrain conditions : forrest, some hills (+/-50m). 

Nothing to claim in my opinion.

Green line: FR945 - track recorded 2 days ago, Red line: FR935 - track recorded 9 days ago.

ok - last post here, until garmin will provide a new fw :-). i had a ran with my old 920 and the 945. ok - i'll admit - the 920 hasn't been started in over one year hence the bad track at the beginning - lesson learned - should have left it a few minutes more to the the gps soaking. anyway, here it goes (945 on the left wrist, 920 on the right, both set to 1 sec, both gps+glonass):

https://connect.garmin.com/modern/activity/4357369076 -945

https://connect.garmin.com/modern/activity/4357389560 - 920

as you can see, the 920 tracked way smoother all U-turns; it went a bit jumpy on the upper track - unfortunatley i remember having to overtake some slower runners there - that might be the case - as the 945 recorded it too. 

the lake - as said - not enough gps soaking :-), as when it returned the 920 tracked nicely. the sudden 90 degree angle is due to a stop :(, the 945 recorded it too, albeit not that visible. so overall i still like the 920 tracking more and i hope garmin could do something to improve accuracy in the 945, to get it on par with the 920/935...

merry christmas everybody,

tudor

ps: oh how the 920 looked and felt like a brick :-)

The difference in the tracking of the path is different but the difference in distance etc was what I would call trivial. There’s 37m difference between them with the 945 tracking 37m further and 2m higher in elevation gain. I don’t really care what my track looks like on a map as long as it follows pretty closely where I’ve been. It’s more important that the distance metrics are accurate. Next question, how long is the actual course? Has it been measured with a calibrated wheel? If you have done this run on several occasions, what are the distances that have been recorded? How much variation is there?

Unfortunately I think that they can't get it to 935 level in software, I think it's crappy at the hardware level. Low power, low accuracy :( 

But as everyone is using Sony's chipsets now, I would hope that Sony will release better ones, maybe some that has adjustable power usage which can adjust the power/accuracy.

I think the old chipset was: https://labs.mediatek.com/en/chipset/MT3333

Power consumption (GPS+GLONASS)

• Acquisition: 37 mW

• Tracking: 27 mW

And the new Sony is probably CXD5603GF so https://www.sony-semicon.co.jp/e/products/lsi/gps/product.html 

So with this, it uses really less power, but we didn't get so significant improves to the watch tracking time, so maybe the new processor in 945 uses the difference or something else. Of it's some best/worse case differences depending on settings, like the refresh rate on MediaTek. But it's easy to see, that if those are the power differences there might also be some accuracy differences.

https://wiki.openstreetmap.org/wiki/GPS_Chipset

There are some differences in possible refresh rate, MediaTek is up to 10 Hz, Sony is 1 Hz, then there is differences in channels, but is that just how they are listed, or is there notable difference how do they work. 

If it looks bad, then the the distance accuracy is based on luck? :) As if it doesn't track where you did run, then how can you think that it's accurate?

If it happens to get good number and looks bad, then it seems to be because of the errors seems to cancel out each others, but not because it's accurate and that feels bad.

hey - indeed - but i would look into a mix of map draw (or actually positioning) and distance on both watches; and the map draw has an impact on the pace - if it tracks erratic - not in a straight line - does zig-zags and so on, that would result in a longer distance and a lower pace (instead 5:00min/km you'll get 5:05-5:10min/km). and if you're doing several loops, if you're placed during one loop with that 37m offset and in the next one with -10m on the other side of the reference line, that difference will stack up with every loop you take (thinking about the song here) :-) . unfortunately the course has not been professionally measured -it's about 2100m and the funny thing is that the 920/935 almost always lap a bit sooner (5-10secs) compared to the 945; that's a thing i've noticed the first time when i wore the 945 as it didn't lapped where i was used to :-). but in order to make the test a bit more professional, i'll follow your suggestion and measure it with a wheel...

if you wouldn't mind, please take a look over the distance recorded on past runs on the same course and look at the variance.

jmto I accept your point about the track placing however, I know from my accurately measured parkrun course that even if the track is 'wobbly' the distance is right. And that, I believe, is due to the processing of the data that Garmin do.

The track display and the distance measured are two different things IMHO. And I'm more interested in the distance metrics...although I do like to see pretty pictures of the track too.

Ok, by luck I mean then "normalization algorithm", which can always have problems :)