All this talk about polar plots, had me looking at today's session.
This is with 3 GPS's, blue on boom, green on head and red on arm.

A 60/40 starboard assy fin, as speed runs were on starboard.
My best speed was very broad, after which I crashed.
Pointing on the assy is way better on starboard, as you'd expect.
But I didn't realise it was as much as this.



It's interesting that the 3 units report different angles, not sure why that is?

(edit) A thought maybe the crash, hand, head and boom going in different directions?

Certainly looks like it here. 5hz device is my arm unit. boom is out there by itself.


Another question is, did the assy fin work better at going broad as well as going upwind?

It's fine if there's room to go upwind on the good side.

decrepit said..

sailquik said.. Fin runs at lower angle of attack to better Lift to Drag ratio. Board runs straighter with less crab. This also means you can run a smaller fin less drag again. Win Win!

Until you gotta get back upwind.

I didn't put that well, I meant does the assy allow you to go deeper than you would with the same fin with standard foil

In as much as you would have slightly lower drag and higher efficiency, you may gain a few degrees, but the limit on how deep you can go for max speed is more controlled by the true wind speed.

That is: the stronger the wind, the deeper you go for max speed.

Its very hard to see how your body could actually accelerate faster than you were going before the crash. Usually, it would be the board catching, or slowing and your body continuing at the previous speed. Although we call it a 'catapult', there is little chance of actually gaining more speed than your were sailing at unless you very vigorously launched yourself forward with you legs? I cant really see something like that happening. Your body mass would need a LOT of energy input from somewhere to accelerate it, and even if that somehow happened, it would be for a very brief time.

The most likely reason for a spike in your speed data is a artefact of the GPS losing signal when it submerges after the crash. We have seen instances of where this can create a spike. but that should be pretty obvious on the speed data and graph if it occurred.

The difference in the plotted courses in the example above is more of a mystery to me, although we have certainly seen instances where plotted course can be incorrect when using Course over Ground (Doppler derived) data v's Positional data. (The reason we use positional data for Alphas rather than CoG data) Positional data is 'self correcting' at each successive plotted point, whereas CoG can veer off on an incorrect course if one point is erroneous.

why would it happen to winging and not windfoiling?
Today had a crash about 2secs after max speed, but looking at the points it seems to agree with my experience..

This thread has gone a bit off track, it's not about spikes giving false data, every way you look at it the data looks good. It's just the difference in the 3 GPS's indicates most of the 2s result was when my body wasn't on the board.
I'll see if I can bring it back.

sailquik said..
Its very hard to see how your body could actually accelerate faster than you were going before the crash.

Andrew, it all happened too quick, and I'm not sure what happened, but the board didn't trip. I think I lost control and got pulled out of the foot straps, as far as the sail was concerned, board drag vanished, and I was accelerated.

sailquik said..Your body mass would need a LOT of energy input from somewhere to accelerate it, and even if that somehow happened, it would be for a very brief time.

Yes it was very brief over 3, 10hz points so 300m/s
Here's the data
35.39
35.605
34.919
35.745
35.637
34.519
35.986
36.146
36.348
37.090
36.896
37.160
36.550
36.861
36.861
34.939
34.128
34.311
34.544
34.122

Certainly not a (spike), a 2kt acceleration over about 1sec, then a 3kt deceleration over 0.6 secs.

Consistent to what I remember, going round the side of the board, and crashing in the water.
Also consistent with the 3 tracks, where they start going in different directions just before max speed.

sailquik said..The difference in the plotted courses in the example above is more of a mystery to me,

What's wrong with my idea that board and body are detached, no longer following the same course?

That's what made me start investigating what is going on.
I originally took it for granted the 2s result was while I was still sailing and the start of the crash happened afterwards

Here's the speed graph

OK, let's assume that the time between detachment of your feet from the board and hitting the water was about a second.

Yes, it may be possible if the sail was still pulling hard for your body to be accelerated a bit. But how much?

But Andrew may be onto something with the 0.4 seconds delay from the crash and the speed spike. I think that is more likely than your body accelerating significantly, and may explain part of that peak speed, but when does the possible 'spike' start? Is it fair to assume that you were accelerating hard when you got detached. If so, is it possible that the speed, up to about 35 knots at least, is real? Do we have enough data to support that anything up to 0.4sec,(or any other period) before the error data starts climbing should be discarded?

Now you explain it again, that may also explain the track divergence between the boom and the body GPS's, as the boom would be going in a different direction from your body. So how much have you zoomed in on the end of the track here? That info could explain the difference between the boom and body units at the start of this picture.

Do all the GPS's give an almost identical speed graph?

It's hard to discount any of these theories, but it is likewise very hard to confirm them.

If you would like to send me the files Mike, I will try to get a scale on it in GPS-Results with the 1m grid.

Sorry for the reply delay Mike. I got a bit distracted playing around with the really interesting new features in GPS-Speedreader (see other thread).

I think this screenshot explains most of it:



In the Trackmap, at this scale, you can clearly see the slight positional separation on the other runs, of the boom mounted GPS compared to the two Body mounted units.

In the crash, just after the best 2 seconds, you can see the boom and body diverging. I am assuming this is when you released the boom and went your seperate way. (Which means it would be impossible to see how your body could accelerate at that point)

There does not seem to be any 'spike' in the speed just prior (in the half second) to you stopping and no evidence of the GPS's completely losing signal (The error does not go above 1.5 on the boom unit and not above 1 on the other two), as we have sometimes seen in other crashes).

The track indicates that the best 2 seconds was just under 1 second from before you hit the water, and couple of seconds before a complete stop. This seems like a fairly normal sudden stop from the data with no strong evidence there was any unusual spike. (In this case it's great that we have a lot of fine data at 10Hz.)

For what it's worth, I would find it very hard to argue that the crash caused any artificially high reading on your GPS units in this case, especially the red and green, which correlate very well.

tbwonder said..
I have been looking at quite a few tracks over the last few days from our new winging friends. Quite often the maximum speed for the session is being reported just before a crash. Which is pretty common when winging. It seems often that the reported speed increases after a crash and unfortunately the error data takes a second or two to reflect this. KA72 is particularly bad at picking this up but even GPS Speedreader (below) has a problem with this issue. Perhaps GPS Speedreader could look at the error data 2 seconds after the best 2 sec run and discard the run if the error data goes high.

In the examples below the speed data is not rejected until the error data goes above 1.2kts (Default)

Question: Do you still see theses 'spikes', if the error does not go above 1.2-1.5?

ozzimark said..
Food for thought: Are we ignoring the vertical axis? Sure, the sail isn't accelerating his body horizontally anymore, but gravity certainly is. a^2 + b^2 = c^2 and all...

Possibly.
How much vertical acceleration would there be when the body drops less than 1m over a horizontal distance of a few meters?

And would that vertical acceleration exceed the horizontal deceleration? If so, by how much?

Mike original question was

decrepit said..
It's interesting that the 3 units report different angles, not sure why that is?
(edit) A thought maybe the crash, hand, head and boom going in different directions?

Quite likely. Maybe post a screen shot of the data, with angle to wind, speed, and error showing?

sailquik said.. >>>For what it's worth, I would find it very hard to argue that the crash caused any artificially high reading on your GPS units in this case, especially the red and green, which correlate very well.

That's what I said from the begining, but leaves unanswered the original question of why max speeds are at different angles on the polar plot. And if that infers, I'm not connected to the board at the time.

sailquik said..
Possibly.
How much vertical acceleration would there be when the body drops less than 1m over a horizontal distance of a few meters?

And would that vertical acceleration exceed the horizontal deceleration? If so, by how much?

Getting a bit sidetracked here, but this was interesting to me.

By definition, horizontal deceleration due to drag would be the same as gravity at terminal velocity, which is quite fast compared to what we can do with wind power. Drag is of course squared with velocity, so I'll assume that deceleration is equal to 9.81 m/s^2 * (horizontal velocity / 90 knots)^2

fall time = sqrt ( 2 * height / g)
vertical velocity = acceleration * time
horizontal velocity = initial velocity - drag deceleration * time

Starting with 30 knots, he'll slow down at first until the vertical velocity component becomes large enough to start increasing the total velocity vector magnitude.


If we assume he's a bit slower at the start:


Or a bit faster...

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?
If so it is an incredible acceleration. I had always assumed that spikes like these were caused from errors within the GPS chip. Generally to do with loss of satellite reception.

Either way I think these spikes should be filtered out in software.

I don't believe a 2 sec peak should be allowed in Team challenge if it is adjacent to unacceptable error data, personally I think there should be 2 seconds of clean data after the 2 sec peak.

If this was submitted to Team Challenge admin as a 2 sec PB I assume they would reject it, even though it has passed the default filters in GPS Speedreader.

tbwonder said..


Either way I think these spikes should be filtered out in software.

I don't believe a 2 sec peak should be allowed in Team challenge if it is adjacent to unacceptable error data, personally I think there should be 2 seconds of clean data after the 2 sec peak.

If this was submitted to Team Challenge admin as a 2 sec PB I assume they would reject it, even though it has passed the default filters in GPS Speedreader.

2 seconds before the error spike may not be wise or workable. It may have rejected Decrepts recent 2 sec result, which I believe is legitimate.

How would you suggest we define 'unacceptable error' in this case?

tbwonder 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?
If so it is an incredible acceleration. I had always assumed that spikes like these were caused from errors within the GPS chip. Generally to do with loss of satellite reception.

The question is if the observed speed increase is real or an artifact. One reason I assume it is sometimes real is that I often feel quite a bit of acceleration when I leave the board behind in a wing crash and keep flying a bit before hitting the water.

Until now, I have assumed that the satellite number would go down quickly when the GPS gets submerged in a crash. That's based on some anecdotal observations, but may not be correct. I just did a few experiments dunking the GPS into a water bucket, and it seems to take about 10-15 seconds for the satellite count to go zero. At the beginning, it still reports an unchanged number of satellites. Once the number of satellites goes to 0, the GPS freezes the reported speed for 10 seconds before dropping to 0.

tbwonder said..
I don't believe a 2 sec peak should be allowed in Team challenge if it is adjacent to unacceptable error data, personally I think there should be 2 seconds of clean data after the 2 sec peak.

While this seems reasonable at first glance, there are multiple problems with this. Perhaps the biggest one is that it will not affect any posts from ka72.com, since changes to the speed analysis algorithm there just will not happen. Fixing a previously reported threshold bug never happened (AFAIK), and that would have been a much easier fix.

The second big issue is relevance. This seems to primarily affect foil data, and in particular beginner/intermediate crashes on slow foils, and in just one category that is known to be the least accurate. It does not affect monthly rankings, unless a team has only foil posts - but then, they'll be at or near the bottom of the rankings, anyway.

A third issue is that this relies on some assumptions about error data which may not always be correct. When finning, it is quite possible that a real top speed in a big gust is followed by a crash. When I hit 40 knots the first time, it was immediately followed with a huge wheelie (tail walk) that could have easily been a huge crash. I would regard not "allowing" a real top speed because it was followed by a crash as totally unacceptable.

I understand that wingers may want accurate results. Anyone who uses GPS Speedreader can easily check the tracks, and simply delete any false top speeds with a couple of mouse clicks. But for most wingers, supporting track analysis from watches that do not have any accuracy data is much more relevant than some small improvements for u-blox based GPS units.

I actually think that automatic crash identification in u-blox data is an interesting problem that probably can be solved. Currently, my best guess is that we need to look at satellite CNO numbers, which seem to go down immediately, a long time before error and satellite numbers change enough to trigger filters. But this will need some further investigation.