GPS-Sport ForenRun.GPS Softwarespeed smoothing
Bitte einloggen, um neue Kommentare zu erstellen Foren durchsuchen
Erste Seite<Seite1von1>Letzte Seite
tomgps 04.09.2007 22:37:25 UTC

speed smoothing in GPS trainer UV has improved, but can be better.

In the new version 2.0.1, speed smoothing works different from 1.2.8. Better, I think, but it is not perfect yet.

the dynamic altering of buffer size is still there (good) but it is less agressive then in 1.2.8. It varies between the buffer size set in the configuration screen-> general settings when riding at constant speed and half of this value when speed varies. In 1.2.8 it varies between 5 sec and the set buffer size. In 2.0.1 the formentioned behaviour of jumpy average speed is solved. When walking with a buffer size of 60 sec, the buffered speed now varies nicely between 5 and 6 sec. At higher speed this variation is even much better even.
Still, I think I found two flaws in speed smoothing of 2.0.1:
1) not sure yet, but it seems that with the smallest possible buffer size of 5 sec the smoothed speed is somewhat higher than the GPS speed. I tested this during car driving it speed of about 90km/h
2) independant of set buffer size, the speed smoothing after a sudden stop is too long. It seems that it always takes 20 sec for the smoothed speed to reach zero after a sudden stop. The drawback of this: a real stop of less than 20 sec is not detected as a pauze, so the Active time is a little too high and the Avg Speed active is a little too low. During a sudden accelaration from a stop this does not happen: smoothed speed is increased from zero within 4 sec, independant of buffer size.
The advantage of this phenomenon: signal loss during less than 20 sec is not counted as a detected pause. Not sure about that yet. I will report on pauze detecting and signal loss in another thread.

All in all: I am happy with the changed behaviour in 2.0.1

Tom
tomgps 17.08.2007 22:05:00 UTC

some more findings on speed smoothing

Today I found out there is no mixup of the buffered and unbuffered speed in the program. When slowing down for a traffic light the unbuffered speed (gps speed in the training screen) goes to zero at the stop, and the buffered speed ((speed in the training screen) goes to zero more smoothly, just as expected from an averaged speed.

But what is the error, then?
In the Training screen, between Speed and GPS Speed the current buffer size is shown. It looks like the buffer size is changed dynamically during the session. For the greater part it is 5 seconds but under certain circumstances it can rise gradually to a maximum of the manually adjusted Buffer size (in Configure--> Speed). One of the circumstances under which the buffer size rises is sudded acceleration and decceleration (like stopping at the lights). I found out that the forementioned behaviour of jumpy speed reading does not occur when the dynamic buffer size is greater than 5 sec. Then the speed shown becomes nice and smoothed. Smoother then the unbuffered GPS speed, like it should be.
My theory: it is a software bug that is introduced by the dynamic altering of the buffer size.
As a user of RunGPS you might think: I am not interested in usage during driving, but this behaviour is even more marked when cycling, running or walking.

Can one of the authors of RunGPS shed some light on this?

Tom
tomgps 15.08.2007 22:05:07 UTC

Did some tests on speed measurements on 1.2.8

Driving in my car on the highway: no trees, excellent GPS signal.
I try to keep my speed as constant as possible. My unbuffered (unsmoothed) speed in the training screen stays between 93.0 and 95.0 km/h. But my smoothed speed (60sec buffer, although the training screen shows 5sec) varies between about 86 and 102 km/h (7% variation) within a few seconds. I tried with lower speed and the variation in the smoothed speed still is about +/- 7%. The speed graph seems to be based on the smoothed speed and thus shows this unrealistic speed variation.

I don't understand this. The smoothed speed should show less variation. I noticed the same at lower speed when running. Seems like a bug to me. You didn't by accident mix up both speeds in the program??

Tom
Erste Seite<Seite1von1>Letzte Seite
© 2024 | Impressum | Allgemeine Geschäftsbedingungen | Nutzungsbedigungen | Datenschutzerklärung | Widerruf und Rücksendung | Batterieentsorgung