New Altimeter algorithm not working, firmware is 20.30

When connected with a phone a combination of Garmin Connect and a course phone location (Google) would not require GPS to be turned on. So why are they not using that to get a GPS 'fix' then?

punkgeek said:

I think a lot of people misunderstand the fundamental physics of how an altimeter works. 

Sorry to say but I think that it is you who misunderstands. In general all of those that I know have valuable input in these discussions are well aware of how altimeter works. The topics are about changes in the altimeter behaviour from the previous stable firmware and the latest in which Garmin very clearly says that there are changes in the "algorithms" for calibrating the altitude. This includes how changes in pressure influence the altimeter throughout the day.

What everyone (Garmin, Polar, Coros, Suunto,...) does is to make decisions if pressure changes are natural or changes in elevation based on context and data from other sensors. What we are saying is that the latest changes to the software making those decisions are not working.

As others say, if it worked better with some other algorithm and now it works worse, then it's not due to physics, given that physics shouldn't have changed in the meanwhile.

I agree with you than inside a house GPS is not activated. The problem is that altimeter works bad even during activities, always closing loops that end tens or meter higher.

I bought this watch because it was the only watch with cartography on the market. Or do you think that, just by having bought it, I must be content with its performances?

Just adding that I have the same issue. The altimeter on 19.20 was bad (so bad I wished I could just disable it and get connect to just automatically enable elevation correction...). When I updated to 20.30 public, with no beta in between, I did not realise it would get worse! Sometimes I manually enter my elevation and before I have even pressed the back button, it has already drifted again.

I think part of the problem is that the watch now no longer does a good job at choosing altimeter vs barometer in auto. The regular drifts in activities, sometimes changing elevation gain by hundreds of metres was present before. But at least when I was sitting on my couch the altimeter was constant (you know, not that this is what I want an altimeter for....).

I hope they fix this and don't neglect it with the f7 out....

PS: I am not in Sweden, just as that was added as a suspicion. (Although at Northern latitude)

claes.bostrom said:

I am trying to find out how other people managed to get it to work.

• 1. Installed the update 20.30 - (as an aside I've always installed any beta software).
• 2. Calibrated with known manual elevation before bed - (44m)
• 3. Went to bed with watch in Battery Saver Mode and option to have Phone Connected..
• 4. Woke up, elevation at 44m
• 6. Randomly checking elevation over the next few days and it's still working fine - effectively left alone as I don't need to keep checking the elevation every 5 minutes.For instance at home elevation is 44m, work is a few meters above sea level, watch is currently showing 8m.

What I have never done - backdated software from a beta, RC, or public release version to an earlier version and/or tried to instal earlier versions of any other sensors.

Elevation profile from a 'flat' run this morning (36m gain) shows a small difference between start and finish:

And a slightly lumpier loop (350m) from a few days back showing a similar small difference between start and finish.

5520171 said:

When connected with a phone a combination of Garmin Connect and a course phone location (Google) would not require GPS to be turned on. So why are they not using that to get a GPS 'fix' then?

Because google course location info will not be able to correct for a 20 m change due to airpressure (or anything else because the cell tower based positioning is even worse than the (crummy) GPS elevation data).

I think people worried about a 20 m change (i.e. small) over hours are seeing a properly functioning barometric altimeter ;-).  

I have been running the same route for weeks and the altimeter always determines the same values.

I don't recalibrate it before I sleep, I never recalibrate it.

Ok, lets see what we have in common.

>>1. Installed the update 20.30 - (as an aside I've always installed any beta software).
    I installed the first of the betas, do not recall the actual version.
    And I backed out of it as soon as I discovered the altimeter drift. Lost all settings in the watch...
    
>>2. Calibrated with known manual elevation before bed - (44m)
    I have done this several nights in a a row, does not help.
    
>>3. Went to bed with watch in Battery Saver Mode and option to have Phone Connected..
    My phone and me are in separate bedrooms during the night ;-) so we don't disturb each other. Has been like that for ages.
    That is a difference. Maybe worth looking into.
    
>>4. Woke up, elevation at 44m
    Woke, up, altitude all over the chart, both plus and minus from my base value of 55 meters.
    
    >>5.
    Wonder what you had as number five :-)
    
 >>6. Randomly checking elevation over the next few days and it's still working fine - effectively left alone as I don't need to keep checking the elevation every 5 minutes.For instance at home elevation is 44m, work is a few meters above sea level, watch is currently showing 8m.
    Before the upgrade to 20.xx, I rarely looked at the altimeter, only the barometer, and was never worried about it showing the wrong numbers.I trusted the numbers. After the upgrade, the feeling is the opposite.

Let's see if we can get any more suggestions like this. Thank you for the input.

claes.bostrom said:

I would not think that a 20 meter difference from a base value of 55 meters

That's not how it works. You cannot do this calculation, given that "55 m" is a conventional measure. You could move your baseline (say, not refer to sea level but 1000 m below) and out of magic you will have an error of 20 m out of 1055? No, you must refer to a real physical quantity, as the pressure. You will have like 3 hPa out of circa 1000 hPa, a 0.3% error.