Looking for % grade tuning/accuracy

[trekker]

I had a bike ride that featured climbing a very steep hill.  The speed was very slow.  The % grade reported on playback didn't come close to the actual.   It had numbers as low as 3.6%.   My calculation shows a grade of ~25% (using the altitude from the GPS file and the distance from playback).  I assume that distance is the distance across the surface not the projection onto a flat plane.  

When playing back and stopping at each step/point during the climb, the % grade should at least be monotonically increasing with each point (more altitude) but it doesn't.

Did my very slow speed cause TF to throw out points?  Even slower speed when projected to level, though some climbing was near 1 ft/sec vertical.

[Krein]

Grade calculations on GPS tracks are, in general, not very accurate.  This is mostly due to errors in the elevation dimension, compounded by errors in 2D position.

Believe it or not, we already do some amount of smoothing to get more accurate results.  It was even worse before.

The grade should not necessary increase as you progress on a climb.  The actual elevation number has nothing to do with it.  Climbs can get less steep as they go on, right?

Also, grade is defined with respect to 2d distance, not terrain distance.  Rise/run.

All that said, I'd like to make grade numbers more useful/accurate, too.  Perhaps it's time for an option to turn a more aggressive smoothing algorithm on.

[trekker]

Quote (Krein @ Sep. 07 2006,1:38)

Grade calculations on GPS tracks are, in general, not very accurate.  This is mostly due to errors in the elevation dimension, compounded by errors in 2D position. Believe it or not, we already do some amount of smoothing to get more accurate results.  It was even worse before. The grade should not necessary increase as you progress on a climb.  The actual elevation number has nothing to do with it.  Climbs can get less steep as they go on, right? Also, grade is defined with respect to 2d distance, not terrain distance.  Rise/run. All that said, I'd like to make grade numbers more useful/accurate, too.  Perhaps it's time for an option to turn a more aggressive smoothing algorithm on.

The tuning needs some confidence factor for the elevation.  The precision (accuracy when calibrated) of the barometric elevation of the 76CSx is very good.  With barometric data the smoothing needs to be less aggressive and more trusting!

The comment was referring to starting from level (0%) as one ascends there IS a range in which the average grade will increase (assuming the range of averaging is more than 2 points).  The program's behavior seems close to random.

I agree on grade definition, I computed run by using the distance from the playback as the hypotenuse.   But this must be circular, isn't altitude input to TF's distance calculation?   I suppose I could get run directly from lat/long.

[Krein]

I agree, the Garmin units with barometers are quite accurate.  So I'm a little puzzled as to why the grade numbers are seemingly random on your track?

I just looked through a few tracks and found that the grade numbers in the profile window are better than I remembered.  They seemed quite reasonable, actually.

Still not sure I agree about gradually increasing numbers.  Trails, roads and the earth in general are quick to change steepness.  Even if it doesn't seem like that in real life, at the average spacing of a GPS point the transition can seem "sudden."

?

[brian.r.hamilton]

I've always thought TopoFusion's calculations were very good in this respect.  I use the 3 meter windowing for my GPS data for cycling & 1 meter for hiking data.  The exact grade at any point can only be as accurate as the data you collect.  You need a lot of points to see grade details for a short climb, especially at faster speeds.  I have my GPS set to collect as much data as possible while still in the Auto mode (or whatever it's called).  You may want to check your GPS settings to ensure you are collecting enough points in your tracks to get the accuracy you want.  You may have no choice but to walk the climb if you want really good detail.

I don't see any reason to change the calculations in TopoFusion.  It would be nice if TF saved the default selections in Climbing Analysis between sessions though.