Select to expand quotetbwonder said..
We need a volunteer to intentionally generate a catapult on a 40kt run to see if a higher speed can be attained.
I do not recalling seeing speed artifacts in my speedsurfing crashes. I specifically checked a crash running onto a sandbar where I broke a boom, and the speed in the crash went down rather than up. I did not see it often when windfoiling, either, but do see it in many wing sessions. At the beginner/intermediate level where such crash artifacts are most common, it makes a lot more sense to look at 5x10 numbers than at 2 seconds, anyway, and 5x10 is a lot more robust against all kinds of artifacts. I think even many watches give pretty decent numbers for 5x10 .
A while ago Bender came up with 60kts on his GT31, and spinning eyes, after a massive catapult
Select to expand quotesailquik said.. >What Version are you running? Me = V2.21
I guess it's the same, 2.2.1
I'll send the files to Peter, see what he gets.
As expected, Peter has explained it. It's that old doppler position accuracy thing. Doppler doesn't track position it tracks angles. Any bad data can through the angle calculation out, which gets carried through to further data, The longer the distance traveled the greater the error becomes. (this is why we don't use doppler for calculating the alpha 50m gate). So my tracks where calculated from the start of the session, Andrews version is from a much shorter sample, so is more accurate regarding relative boom and body positions.
Decrepit and I had another look at his files while on the phone tonight and it was interesting. We solved the different trackmap picture issue, but it still left us scratching our heads a bit.
When we trimmed the data to just before and after the crash, and at the start for various lengths, the pictures didnt change, so it appears it was not caused by a heading error close the crash, ot at the star of the session.
But when we applied 'delete filtered points' the picture changed very significantly and i saw the same picture that Decrepit posted. This suggest that the heading errors were in some of the filtered points, although the second (deleted filtered points) graph looks wrong in other ways.
Also, my initial assumption that the more offset track was the boom unit (which would be logical) was incorrect.
So that left a few mysteries that we dont have an answer for.
In any case, we agree that this 2 sec peak is most likely to be quite legitimate, and we have further explored some software capabilities, so we did learn something. 
Here are the two track pictures we get. First without deleting filtered points. Second after deleting filtered points.
(Its relevant to note that the actual 2 sec results didnt change, just the position in the map of the 3 tracks)
.
Boardsurfr put us onto the reason for this, which is deviations caused by errors in the Doppler Heading.(which 'accumulate' - for want of a more precise description)
It would be interesting to be able to superimpose positional track data over this, but although I believe GPS-Results can do that, I have not been able to open these tracks in it together for comparison so far.
Select to expand quotesailquik said..
Boardsurfr put us onto the reason for this, which is deviations caused by errors in the Doppler Heading.(which 'accumulate' - for want of a more precise description)
No, that's not it. The tracks are always based on positional data. In the graph, they are centered relative to the first point in the track. The positional error between devices is different, and varies over time. When you trim points at the start of tracks, a new point will be used as a reference. Other tracks will shift to reflect the new relative position error between the tracks.
If the tracks were based on doppler data, they would likely diverge a lot more over long tracks, since the errors could accumulate - especially any non-random errors. Here is a comparison of the tracks drawing from positions (latitude and longitude) and doppler speed and headings, using the 2 10 Hz tracks from Mike, near the start (after deleting the first 8500 points and erasing filtered speeds):
Near the start, using Doppler data to draw the tracks would work well. But over time, errors accumulate, and this is what it looks near the end of the track:
In the accumulated doppler data, the tracks have separated by about 40 meters in x direction, and about 60 meters in y direction. Meanwhile, the distance between the tracepoints is always less than about 1.5 m, roughly the positional accuracy of the GPS chips.
The reason that you and Mike saw the differences were that one of you looked at tracks with points removed at the start (or you removed a different number of points, if you both trimmed at the start). Where the aligned tracks start (when all 3 units have valid data), the positional error can be a bit larger than later because they are still gathering satellites.
Select to expand quotetbwonder said..
Boardsurfr are you suggesting that in my example posted above (and shown again here) that the spike is caused by the sailor floating through the air after leaving the board?
I had a fun wing session today where I picked a wing to be well powered, and then was nicely overpowered when the wing picked up 5 knots, but I could not go in to switch because it had to be a short session. I have several crashes where I had enough time to think "this feels like nice acceleration!" after leaving the board. In at least one of these, I noticed that I definitely got extra power because I was pulling on the back handle trying to regain my balance.
In my top 2-second speeds, no less than 4 of 5 are right before a crash. Here is one example where I definitely remember feeling a speedup in the crash (it was on my way back near a marker so I know the GPS data belong to this crash):
The speed gain was about 1- 1.5 knots over 2 seconds. My GPS was on top of my helmet, and I was hanging on the wing so it never lost satellite reception, and error values never got close to the filter thresholds.
Here is another example:
This one was going into a jibe and overfoiling. A bit more acceleration here, about 2.5 knots in 1 second. That's only about 1/8th g, so quite reasonable when suddenly reducing drag to 0. Deceleration is about 3x higher.
Those are two examples where I am 99.99% sure that the spike happened when I was flying through the air (in the second example still connected to the board at the very start). The spike in the data you had posted is a bit more pronounced, and the GPS looses reception shortly thereafter (judging by error estimates - satellite numbers can be misleading since they can take a few seconds to drop). So it's possible that your example is an artifact, but it could also be a more violent crash. If you have a ubx file and it has NAV-SAT data, you can look at the SV CN/O levels to tell when the GPS hit the water. CN/O levels drop much quicker than the number of satellites used. That's partly because typical CN/O levels are between 20-40, while the default CN/O cutoff is 6 or thereabouts.
