What's new

GPS accuracy

Ludwig

Speeding Up
Oct 9, 2008
871
0
BTW, has anyone noticed that Tim did 500m more climbing than Fumiki even though Fumiki added Kazahari after Matsuhime while Tim did not? Is any more proof needed that the Garmin site altitude readings suck?

You get the true climbing only by memorizing what the device says at the end of the ride. Once uploaded to the Garmin site, anything resembling reality has been lost...
 
BTW, has anyone noticed that Tim did 500m more climbing than Fumiki even though Fumiki added Kazahari after Matsuhime while Tim did not? Is any more proof needed that the Garmin site altitude readings suck?

You get the true climbing only by memorizing what the device says at the end of the ride. Once uploaded to the Garmin site, anything resembling reality has been lost...

Just FYI, Tim is using the Edge 705 which has map based altitude calculations. The newer 500 and 800 are barometric pressure based and are significantly more accurate. I've always found that what my 800 tells me at the end of the ride and what I get after uploading are very close. The difference is always within 50m or so, usually less.
 
Yes - this is one of the bothersome points of this device. I'm very happy with the 705 as it was a special 'gift' - but in terms of data accuracy - well, it's about on par with TEPCO's radiation measurements. But hey - I'm older and get an altitude handicap, right?

Just FYI, Tim is using the Edge 705 which has map based altitude calculations. The newer 500 and 800 are barometric pressure based and are significantly more accurate. I've always found that what my 800 tells me at the end of the ride and what I get after uploading are very close. The difference is always within 50m or so, usually less.
 
Sorry, guys, but this is BS - where did you get this idea from?? All cycling computers use barometric measurement, including the Edge 705. I can even tell because it moves mostly in perfect synch with my other cycling computer (Ciclosport).

The problem happens when the stored data is uploaded to the Garmin site. I presume the Garmin site recalculates the altitude gain based on every data point saved. If you save in 1 second increments, that's a lot of data points some of which will have a bigger or smaller delta because of small changes in barometric pressure/measurement error.

So let me guess:

- Tim saves in 1 second increments

- Fumiki saves in 10 second increments

- Pete saves in 10 second increments

Tim (and also myself) thus have a lot higher cumulative measurement error than Fumiki or Pete.
 
Ludwig - FYI - I am saving in 1s intervals. I set my unit that way to presumably have more accuracy especially when using an SRM. I can see what you're saying regarding the cumulative error - I wonder if this applies, also, to distance? Also - the posted results are quite different between Garmin Connect and RidewithGPS - so obviously there are some differences in how those sites are processing the raw data. And, yes, the 705 does have a Barometric sensor and according to Garmin it should be calibrating itself every 15min according to map data.
 
Tim - thanks for so far confirming my hypothesis. Maybe this map calibration works with Garmin issued maps. It certainly doesn't work with the UUD maps I use in Japan, or any other maps I have tried outside Japan (but none of these had altitude data to start with). My 705 can sometimes get quite off the map data and does not necessarily return closer to the truth. But equally this can happen with my Ciclosport. It's all a bit random - as it should be.
 
There are three sets of competing 'live' GPS data for altitude:

1. The apparent point height on the electronic map. This is often confusing because the maps don't take account of tunnels.

2. The apparent height from the barometric sensor (when installed). As Tim notes, this is allegedly automatically recalibrated, which explains the 50m jumps I see on my .tcx files when I park and turn off the Edge 800, e.g., while at work.

3. The apparent height trigonometrically-calculated from the GPS satellite constellation. Accuracy for elevation is notoriously less accurate than latitude/longitude.

I don't know whether there's a good answer, other than to ask my legs: How much have I climbed today? Too much.
 
HF Mike, I believe 2 is because of 3, and not related to a map. You can check easily: does your Garmin really calibrate to the exact altitude on the map? In my case it almost never does and can be quite off, especially just after switching it on.
 
How about the option "elevation correction" in Garmin Connect?
This is supposed to use the data from actual map surveys when the GPS data isn't accurate.
I usually see some small changes in elevation data when playing with this option.
 
Yes, Andreas, if you use this option, the the site uses the data embedded in the Bing maps rather than the altitude data uploaded from your device. In some cases, this reduces the amount of climbing. For me it mostly increases it, sometimes crazily so, because I tend to ride through a lot of long tunnels. So it doesn't really help.
 
Yes, Andreas, if you use this option, the the site uses the data embedded in the Bing maps rather than the altitude data uploaded from your device. In some cases, this reduces the amount of climbing. For me it mostly increases it, sometimes crazily so, because I tend to ride through a lot of long tunnels. So it doesn't really help.

Hmm...tunnels, didn't think about that. For sure that would mess up the elevation data completely.
 
By the way - I started a thread on the Training Sites - maybe we could organize a kind of test set and then compare against a control course(s). Like take 3 different courses and all of us run them - then upload into the various sites and compare.

1) Flat Course - no tunnels, like a simple Arakawa Run.
2) Hill Course - no tunnels, perhaps Shiraishi Toge, etc.
3) Hill Course - w/Tunnels, perhaps Kobu Tunnel -> Kazahari -> Ome. Something like that?

Anyway - lets move this to another thread :)
 
Just FYI, Tim is using the Edge 705 which has map based altitude calculations. The newer 500 and 800 are barometric pressure based and are significantly more accurate. I've always found that what my 800 tells me at the end of the ride and what I get after uploading are very close. The difference is always within 50m or so, usually less.

Oops, I was wrong about the 705. Apparently it does use a barometer. ;)

That being said, I stand by the second part of my statement in that what my 800 tells me at the end of the ride is usually very close to what both Garmin Connect and RideWithGPS give me after uploading my data.
 
Pete - no doubt because you are saving in 10-second increments?
 
Im wondering if any of you have setup the Garmin's Set Elevation feature. Also I know that if your unit has been sat inside and then is taken outside and started straight away then you are going to get massive differences as the unit needs to recalibrate.

The way round it I've found is turn on the Garmin and leave it while you get everything ready, preferably if you do it outside so it colls or warms up to the ambient temps.
 
James, you are definitely right, but the problem I was pointing out is not with the device but the site. The device actually works just as it should. It's the site which shouldn't calculate the altitude gain based on 1-second increments if this is what you set the device to, but it does. This gives a big (feel-good) error.
 
The problem happens when the stored data is uploaded to the Garmin site. I presume the Garmin site recalculates the altitude gain based on every data point saved. If you save in 1 second increments, that's a lot of data points some of which will have a bigger or smaller delta because of small changes in barometric pressure/measurement error.

So let me guess:

- Tim saves in 1 second increments

- Fumiki saves in 10 second increments

- Pete saves in 10 second increments

Tim (and also myself) thus have a lot higher cumulative measurement error than Fumiki or Pete.

I would look for the difference here::angel:
a) Devices are very different (705 against 500)
b) For the 705 data (Tim) elevation correction is enabled in Garmin Connect while for the 500 (Fumiki) it's not :eek:.
 
I'm thinking to do an early AM ride up Shiraishi Toge this Sat. If Pete, or anyone else comes, then we could do a good test by starting exactly same place, riding the same line, etc. That ride has no tunnels, a nice gentle long slope until the main climb which is steeper. Maybe do the main climb a couple times with different device acquisition settings as well. Like, I wonder how much the 'line' affects altitude gathering if you are using 1s intervals? For example on the switchbacks, depending on strategy I would take the shorter, steeper path and even though the net gain is the same, the interval gain would be different. As Ludwig pointed out - I think the site is not calculating properly - or using a method that incurs stastical error.
 
I would look for the difference here::angel:
a) Devices are very different (705 against 500)
b) For the 705 data (Tim) elevation correction is enabled in Garmin Connect while for the 500 (Fumiki) it's not :eek:.

Hi Malte, I think you are right regarding b) but not a). I ride a lot with Tom who uses a 500 and me a 705 and during the ride the measurement is almost always the same. So as I said, I don't think it's the devices. It's the site. Enabling or disabling the "elevation correction" (a misnomer because with tunnels it makes things worse!) makes a difference, but even without it there can be big differences, as I showed with the comparison of Tom's and my recording of the very same trip.
 
I parsed my last TCX (Edge500) in Excel and sum up all ascents and compared it with the Garmin Connect (no correction) value. The difference was 1797m (Excel) to 1563m (GC) over 23.720 Samples.

When looking at the individual samples, I saw that the altitude increments between 0 to 1m with resolution of 0.1m (for Interval between 1s - to 10s -- smart recording).

If you then use the TCX data and calculate the Gradient or Speed from that you get very noise data. So you need to do some smoothing. GC and RWGPS will have different smoothing algorithm that's why you end up having different values (when importing TCX Logs).

The Device itself might have the best smoothing filter as it has access to satellite strength data, compass etc.
 
Back
Top Bottom