Over the last few months my sons GPS files have had a lot of high doppler errors. The unit has been set up as per the mtbest website settings and he has tried wearing it in different positions but still giving high error rates.Any advice would be appreciated.

Due to an artifact of how a GPS works **, you do indeed get quite a few doppler errors when you gybe... it is the primary reason why we dont use doppler for alpha categories.

** everything you didn't want to know about correcting GPS errors... in particular Kalman filters: en.wikipedia.org/wiki/GPS/INS

i though that in dec of 2009 it was decided that the GPSTC would use doppler for alpha categories but how large can the doppler error be before the alpha is invalid as i cant find any info on this.

Yes, Mathew, my memory is the same, we used trackpoints for the old GT11s but once the GT31s came out Andrew did some tests, and decided doppler was the more accurate. I've been using doppler ever since, and I think most other people with GT31s have as well.

My 10 best alphas today all had 0 doppler errors, so something is certainly wrong, but I have no idea what constitutes a doppler error, is it a lose of satellite connection?

Yes, we use Doppler for Alphas from the GT-31, but, (getting technical here), we use Location (trackpoint) for the proximity circle. This is possibly what threw Mat out.

GT-31 accuracy is still too inconsistent for a high level of confidence in Alpha results. There are too many things that can, and do go wrong in a U-turn.

We are hoping that the eventual introduction of 10Hz (or higher) units will help to improve things by drastically reducing the potential for compounding errors and aliasing. Better control of the firmware should help as well. My experiments with the GT-35, 5 Hz prototype were very encouraging. Unfortunately, production of that unit was not deemed practical, and my one and only GT-35 now lies somewhere on the bottom of Swan Bay!

As to what constitutes excessive Alpha error, that is a much more difficult question. The first thing to look for is an unusual departure from the 'normal' "V" shape in the turn on the speed graph. If there is a spike or unusual bump in it, it is probably error. You will see this more often in the Trackpoint speed trace which is why we decided to use Doppler in the GT-31.
Secondy, look for a nice even U shape turn on the map view. If it has a point, it is suspect.
Thirdly. If you have GPS-Results, (but not RealSpeed, unfortunately) you can superimpose the SDOP (Doppler error estimate) over the speed graph. If the 2 dotted lines are wide away from the speed line it is suspect. One little quirk with GPS-Results is that you cant view Dopper graph, Trackpoint graph and SDOP together. SDOP is only selectable in Doppler. (With Doppler off you can select to see both Doppler and Trackpoint graph).

Here are a few examples from a recent session:

This one is clearly an error. See the difference between the Trackpoints and doppler and the sharp pointy turn on the map.


Also suspect for the pointy turn shape and Doppler error on the speed graph on the start of the turn:

The same run as above showing the Trackpoint error. This is why we decided the use Doppler with the GT-31.



This one looks pretty good! This is Trackpoint result (to get both on the graph). Doppler was 26.21.



The bottom line, if in doubt, get a second opinion from your team captain, the local GPS guru or one of the tech advisors. I am always happy to give an opinion!

Select to expand quote

decrepit said..

Yes, Mathew, my memory is the same, we used trackpoints for the old GT11s but once the GT31s came out Andrew did some tests, and decided doppler was the more accurate. I've been using doppler ever since, and I think most other people with GT31s have as well.

My 10 best alphas today all had 0 doppler errors, so something is certainly wrong, but I have no idea what constitutes a doppler error, is it a lose of satellite connection?

umm ... depends on your context... the question was about doppler-errors - in this case, doppler error through the gybe as visualised using RealSpeed.


Doppler-errors accumulate -> when travelling in a (relatively) straight line, the error-accumulate rate, is about 2 orders of magnitude less than the trackpoint error. However, once you turn a sharp corner, the error rate of doppler shoots up (due to doppler-angle errors, not doppler-speed).

So the error-at-the-end-of-a-run in a straight-line is only about 1 oom larger than trackpoint - still better, but not much... For alphas, it turns out that for a proximity-circle, trackpoints have lower error **.


As Andrew has shown, the actual doppler-angle error, depends on quite a few environmental factors... in short, the doppler-speed should still be mostly accurate.


** Manfred Fuchs wrote a white-paper on doppler-angle calculations (as opposed to doppler-speed used in speedsailing) - it was a very interesting piece of work.