Progress! I've figured out how to do a time sort, so the next lot are in time order not speed order, makes aligning the runs much easier. And I can export to png instead of doing a screenshot, so lets see what it looks like on this page!





So that's a bit easier to read, interestingly the earlier alphas when I forgot to use overhand grip are generally the best, and I'm sure conditions didn't drop off. Some of the runs I remembered a bit late and the transition is evident half way through the run in, and I've put a "?" against them. here's one where I changed just in time.

I appreciate all the effort you guys are putting into verifying the new watch.
It is certainly a very cool looking watch and very easy to read.

At Daffy's suggestion, next trick will be to see what happens with both units on my head.




Not sure if the watch will significantly shadow the GW52 though.

I only underhand grip with my front hand on starboard tack. It just doesn't feel right on port tack. I have no idea why, maybe I have restricted rotation of my left arm. The good news is that if I wear the watch on my left arm, I'll always have an overhand grip.

Select to expand quote

John340 said..
I only underhand grip with my front hand on starboard tack. It just doesn't feel right on port tack. I have no idea why, maybe I have restricted rotation of my left arm. The good news is that if I wear the watch on my left arm, I'll always have an overhand grip.

Fantastic John!!!!!!!!!!!!!

Watch on head today, wasn't much of a session but I have enough runs to confirm the watch works much better on the head, although I'd be very surprised to see anybody else wearing it there!

Here's the speed graphs of my best gybe today

Here's the watch on my head.


The SDoP envelope is there, but so small it's almost invisible.

And here's the GW52, possibly with a bit of shadow from the watch


It certainly doesn't look as clean as the watch

Select to expand quote

sailquik said..
Here you are going on about this so called 'noise' again!! Totally incorrect. it is simply data.

As I have stated a bunch of times before, I make my living by writing scientific data analysis software. I have build a successful company around software that I have developed. A large number of the algorithms in the software specifically deals with handling noise - errors in the data from random and non-random sources.

Funny thing is that you are contradicting yourself, although you probably don't see that. One one hand, you absolutely insist that GPS devices provide error estimates like SDOP, which are measurements of noise. But when I refer to noise that is apparently specific to GW-52 measurements (largely re-stating what decrepit said), it's "totally incorrect".

Data = signal + noise.

We want the signal. We like noise estimates like SDOP so we can get an idea how close the observed data are to the true signal.

The alpha curves decrepit posted above show quite a bit of noise, especially after the jibe, which differs between the 2 units. The GW-52 has larger spikes, and the corresponding larger error estimates. If you pick 10 points for a 2 second speed, chances are that the higher noise peaks in the GW-52 data contribute to the higher observed measurements. The more of them you have, and the larger they are, the more likely it is that the 2 second speeds are inflated.

Select to expand quote

decrepit said..
Watch on head today, wasn't much of a session but I have enough runs to confirm the watch works much better on the head, although I'd be very surprised to see anybody else wearing it there!

Here's the speed graphs of my best gybe today

Here's the watch on my head.


The SDoP envelope is there, but so small it's almost invisible.

And here's the GW52, possibly with a bit of shadow from the watch


It certainly doesn't look as clean as the watch

Does that mean I have to buy another watch now. One for my head and one on the wrist

Select to expand quote

Steve Charles said..
Does that mean I have to buy another watch now. One for my head and one on the wrist

Steve, if you check the Locosys manual that came in the GW-60 box the fine print explains that if you only have one watch then the preferred technique is to sail as normal but as you approach your 2S max speed you take your watch hand off the boom and rest it on top of your head. Problem solved. Its one of the few windsurfing techniques that you can practice anywhere (at home, work, while driving, etc).

Select to expand quote

TGale said..

Steve Charles said..
Does that mean I have to buy another watch now. One for my head and one on the wristwww.seabreeze.com.au/images/forums/icon_smile_cool.gif' />

Steve, if you check the Locosys manual that came in the GW-60 box the fine print explains that if you only have one watch then the preferred technique is to sail as normal but as you approach your 2S max speed you take your watch hand off the boom and rest it on top of your head. Problem solved. Its one of the few windsurfing techniques that you can practice anywhere (at home, work, while driving, etc).

WOW.. can't wait till they develope a head band GPS you can wear as a watch

Select to expand quote

TGale said..
>>>>
Steve, if you check the Locosys manual that came in the GW-60 box the fine print explains that if you only have one watch then the preferred technique is to sail as normal but as you approach your 2S max speed you take your watch hand off the boom and rest it on top of your head. Problem solved. Its one of the few windsurfing techniques that you can practice anywhere (at home, work, while driving, etc).

Hmmmmmmmmmmmm, well it would have to be incredibly fine print, as the manual in the box is so small you can't see it.

Select to expand quote

decrepit said..
Watch on head today, wasn't much of a session but I have enough runs to confirm the watch works much better on the head, although I'd be very surprised to see anybody else wearing it there!

Mike, with it on your head can you hear the noise?
Maybe we will need earplugs if we wear then on our head

Select to expand quote

AUS 808 said..


Mike, with it on your head can you hear the noise?
Maybe we will need earplugs if we wear then on our head

With it on your head there's virtually no noise, so all's good.

All you need now is a Blue toothed water proof HUD for Gath Helmets

Had the watch on my right wrist today, If you don't like doing the overhand thing wearing it on the trailing hand into the gybe, is a bit better that having it on the leading hand into the gybe.





As the watch is overhand all the way through the gybe and part of the way out, I only change to overhand when fully up to speed and the sail is raked back.

And the 2s with watch on trailing hand, is very similar to GW52 on head.

Watch




GW52




so the speed is within 0.012kts and the +/- within 0.006kts.
I'll compare all the numbers later. But this is the obvious way to go, it only becomes a problem when there are speed runs in both directions, guess you just have to try and pick which way will be the best.

Processed todays 10s results, I think because I had the watch on my trailing arm they are very close.

Select to expand quote

boardsurfr said..

sailquik said..
Here you are going on about this so called 'noise' again!! Totally incorrect. it is simply data.

As I have stated a bunch of times before, I make my living by writing scientific data analysis software. I have build a successful company around software that I have developed. A large number of the algorithms in the software specifically deals with handling noise - errors in the data from random and non-random sources.

Funny thing is that you are contradicting yourself, although you probably don't see that. One one hand, you absolutely insist that GPS devices provide error estimates like SDOP, which are measurements of noise. But when I refer to noise that is apparently specific to GW-52 measurements (largely re-stating what decrepit said), it's "totally incorrect".

Data = signal + noise.

We want the signal. We like noise estimates like SDOP so we can get an idea how close the observed data are to the true signal.

The alpha curves decrepit posted above show quite a bit of noise, especially after the jibe, which differs between the 2 units. The GW-52 has larger spikes, and the corresponding larger error estimates. If you pick 10 points for a 2 second speed, chances are that the higher noise peaks in the GW-52 data contribute to the higher observed measurements. The more of them you have, and the larger they are, the more likely it is that the 2 second speeds are inflated.

Most of what you said is accurate - sort of.... SDOP isn't a measurement of noise -> it is an summation of the least-squares worst case error that is possible from all known error-producing sources.

Importantly in this context of an Electronic device, the term "noise" is almost always considered to be in the ElegEng definition -> so it almost always is from a "random" source - if it isn't random, then in ElecEng it generally isn't considered a noise source -> it is instead called "bias". [ Bias can usually measured and accounted for. ]

Which brings me to the last paragraph - it is mostly wrong due to the randomness of the noise source -> you are equally likely to get an "all high" reading as you would an "all low" reading... but mostly, you _will_ get "somewhere closer to the middle, than the extremes" and the more samples you have, the more likely you are to be in the middle.

Aside - Andrew's statement isn't contradicting itself - it is also correct. [ Taking the single sentence was taken out of context of the discussion-thread, it indeed sounds wrong... but we this thread has been running for some time -> it should be taken in complete context. ] Having data updated more frequently, just means we have more data... nothing more, nothing less.

I promise I'm not trying to stir up a hornets nest, but I was just reading the thread on the KA Race 2017 and Spotti's session at the Pit last month...
www.seabreeze.com.au/forums/Windsurfing/Review/KA-Sails-2017/?page=2
He was wearing 3 GT31's and all 3 gave different readings for his 2sec peak as shown in his pics. The difference between the fastest and the slowest is a fair bit, easily as much as the differences being shown between watches and GT31's as per the arguments that have been raging all along. If there is such a variance even with our 'industry standard' device, how can we be so sure of even a GT31's accuracy? Surely this brings the watches back into play now as a viable alternative....

As an aside I have been wearing my GW60 purely as watch for almost a week and I VERY happy to say it is still showing FULL battery charge. Admittedly I have to push a button to read the screen to get it to show up but that's no issue. I am really happy with it and look forward to posting a session on the water from it soon. Since I had my GT31 stolen and had no alternative than to buy a GW60 I am obviously no longer able to post to the GPSTC cause I don't have a valid device

Rex, it's hard to comment unless we know exactly where he was wearing the GPSs. It's hard to get 3 GT31s close together all with good skyview.

That's why the SDoP data is so valuable, it lets you know the estimated +/- value of any result. And why it's best if the GPS has a good skyview.
So if there are differences between multiple devices worn on the body, it's easy to check which of them has the most chance of being correct.

I think as a watch asleep the battery should go for months.
I've currently got my GW60 set to 1hz doing a geostationary battery test to see how long it lasts in GPS mode.

I'll be back later with the result.

Select to expand quote

decrepit said..
Processed todays 10s results, I think because I had the watch on my trailing arm they are very close.

Woops, just realised the first run at 13:11:59 was on starboard tack with an underhand grip, that's probably why it has the greatest difference to the GW52.

Select to expand quote

decrepit said..
Rex, it's hard to comment unless we know exactly where he was wearing the GPSs. It's hard to get 3 GT31s close together all with good skyview.

That's why the SDoP data is so valuable, it lets you know the estimated +/- value of any result. And why it's best if the GPS has a good skyview.
So if there are differences between multiple devices worn on the body, it's easy to check which of them has the most chance of being correct.

I think as a watch asleep the battery should go for months.
I've currently got my GW60 set to 1hz doing a geostationary battery test to see how long it lasts in GPS mode.

I'll be back later with the result.

Hey Mike,
if you look at the photo of Spotty it's actually very easy to see where all his devices are. In a pouch on his chest facing outwards.

Select to expand quote

decrepit said..
That's why the SDoP data is so valuable, it lets you know the estimated +/- value of any result.

Mike,

What is an acceptable error range (i.e. +/- value)?

Select to expand quote

John340 said..

decrepit said..
That's why the SDoP data is so valuable, it lets you know the estimated +/- value of any result.

Mike,

What is an acceptable error range (i.e. +/- value)?

That's for somebody with Daffy's experience to to answer, I'm just doing the tests and picking this stuff up along the way.

It varies with time, the longer times have better accuracy as there's more chance for extraneous stuff to cancel out.
But personally I'd be very unhappy if a 10s result was +/- 1kt, typically with a good signal I'm getting less than 0.2kt

With the underhand grip some 10s are over 0.3kts
And some 0.2sec maximum peaks are getting up to 0.9kts.
This is obvious using the GW52 and GW60, you can get vary excited by the peak speed reading only to find the 2s average can be almost a knot lower.

Select to expand quote

sick_em_rex said..
>>>>
Hey Mike,
if you look at the photo of Spotty it's actually very easy to see where all his devices are. In a pouch on his chest facing outwards.

Well not an ideal spot, but if they're all together, with the same sky view, I would expect similar results.

Had a strange one today, checked after approx six and a half hours and battery was showing no bars but I still had a display, went back half hr later and no display, started charging and checked the file only to find it's empty!!!!

I had it in the window, it looked like it had signal, would go above the zeros every now and again. I just checked and logging is set to 1hz.

The GW52 ran out of battery on my distance attempt after 4.95 hrs but all of it was stored, so I expected to see 6.5hrs to 7hrs worth of data.
Better do this test again, maybe with it on my wrist.

Select to expand quote

mathew said..

boardsurfr said..

sailquik said..
Here you are going on about this so called 'noise' again!! Totally incorrect. it is simply data.

As I have stated a bunch of times before, I make my living by writing scientific data analysis software. I have build a successful company around software that I have developed. A large number of the algorithms in the software specifically deals with handling noise - errors in the data from random and non-random sources.

Funny thing is that you are contradicting yourself, although you probably don't see that. One one hand, you absolutely insist that GPS devices provide error estimates like SDOP, which are measurements of noise. But when I refer to noise that is apparently specific to GW-52 measurements (largely re-stating what decrepit said), it's "totally incorrect".

Data = signal + noise.

We want the signal. We like noise estimates like SDOP so we can get an idea how close the observed data are to the true signal.

The alpha curves decrepit posted above show quite a bit of noise, especially after the jibe, which differs between the 2 units. The GW-52 has larger spikes, and the corresponding larger error estimates. If you pick 10 points for a 2 second speed, chances are that the higher noise peaks in the GW-52 data contribute to the higher observed measurements. The more of them you have, and the larger they are, the more likely it is that the 2 second speeds are inflated.

Most of what you said is accurate - sort of.... SDOP isn't a measurement of noise -> it is an summation of the least-squares worst case error that is possible from all known error-producing sources.

Importantly in this context of an Electronic device, the term "noise" is almost always considered to be in the ElegEng definition -> so it almost always is from a "random" source - if it isn't random, then in ElecEng it generally isn't considered a noise source -> it is instead called "bias". [ Bias can usually measured and accounted for. ]

Which brings me to the last paragraph - it is mostly wrong due to the randomness of the noise source -> you are equally likely to get an "all high" reading as you would an "all low" reading... but mostly, you _will_ get "somewhere closer to the middle, than the extremes" and the more samples you have, the more likely you are to be in the middle.

Aside - Andrew's statement isn't contradicting itself - it is also correct. [ Taking the single sentence was taken out of context of the discussion-thread, it indeed sounds wrong... but we this thread has been running for some time -> it should be taken in complete context. ] Having data updated more frequently, just means we have more data... nothing more, nothing less.

Im not sure who is arguing what here but its noise isnt it as a bias would need a direction ? like +0.5 not +- 0.5.

Had lots to say but ill sum it up, your all over thinking it, they all look great.

If they just ran a filter that chopped off the spikes, averaged the last few readings and altered the sdop accordingly you would all think its the best gps on earth and be really happy with it :)

Select to expand quote

TGale said..

Steve Charles said..
Does that mean I have to buy another watch now. One for my head and one on the wrist

Steve, if you check the Locosys manual that came in the GW-60 box the fine print explains that if you only have one watch then the preferred technique is to sail as normal but as you approach your 2S max speed you take your watch hand off the boom and rest it on top of your head. Problem solved. Its one of the few windsurfing techniques that you can practice anywhere (at home, work, while driving, etc).

Well, tried it out today, I did an exhaustiveng single test fairly underpowered in 0.3m chop and compared my two best 2S results. My GT31 was on my head for both results, and for the first result had my GW-60 hand on the boom and for the second result had my GW-60 hand resting on my head (i.e. executing what we could come to call the GW-60 salute):

Test GT31 GW-60 Difference (knots)
Normal 26.300 26.242 0.058
GW-60 salute 25.503 25.575 0.072

Comments:
- It certainly added an interesting twist to getting a 2S peak
- I think I accidently held my hand over the GT31 during the salute
- It seems to be of no benefit whatsoever.

Conclusions:
- be patient
- keep both hands on the boom
- leave the GW-60 testing to the experts (thanks decrepit and others).

Select to expand quote

mathew said..

Which brings me to the last paragraph - it is mostly wrong due to the randomness of the noise source -> you are equally likely to get an "all high" reading as you would an "all low" reading... but mostly, you _will_ get "somewhere closer to the middle, than the extremes"

That would be true if (and only if) either the 2-second peak would be very defined, or we would be taking a _random_ 2-second sample. Neither is the case.

The speed is often quite close to the maximum for longer than 2 seconds. For example, assume that the true speed drops by 0.2 knots right after the 2-second maximum, but we have a +1 knot random spike there. This will cause the measured speed to be 0.8 knots too high. Since the software is searching for the 2-second region with the highest speed, it will select this peak, and overstate the speed. Because we are actively looking for the 10 data points with the highest average, we are much more likely to pick up a random peak, and overstate speed.

Note that I am not talking about an "all high", but rather about an "edge peak" effect. Even if 6 of 10 points are accurate, and 4 points have a +1 knot error, we'll overestimate the 2 second speed by 0.4 knots. The noisier the data are, the bigger the over-estimate will be. That's exactly what decrepit's 2 second results showed. The GW-52 had higher SDOPs, caused by more noise, than the GW-60 data, and the 2 second speeds were higher. Due to random noise.

Here's a simple thing to do to see the effect noise on searching 2-second maxima: make a little spreadsheet with a few hundred random data points between -1 and +1. Add a column where you calculate the average over 10 points. Then, search for the maximum in the averages over different ranges. As soon as you look in 100-200 points, you'll get an average of close to 0.48. Shorter ranges will give you lower numbers, but even looking at just a 3-second range will give a bias of 0.16.

Bottom line is that noise will often lead to inflated speed values, and that the effect will be proportional to the amount of noise. Higher SDOP values indicate more noise. In decrepit's 2 second data, the GW-52 had higher SDOP values and thus more noise, and also had higher speed estimates. Exactly as expected. The difference in the 10 second values is a lot smaller, since the contribution of "edge noise peaks" to the average is 5x lower.

Nothing new here, really. We've seen the same thing with other less accurate GPS devices in the past, which is one reason the 5x10 number is so popular. Why Daffy got so excited because I use the word "noise" escapes me.

Since I like noise to much, I played around with it a bit more, and blogged about it at boardsurfr.blogspot.com/2016/12/fun-with-noise.html

Figured out what I did wrong on yesterday's battery test, I still had "min speed" set to 2kts. As the device never moved, nothing was recorded, silly old fart moment!
So today I set min speed to 0kts , even took it for a drive into Mandurah, signal isn't too bad on my wrist, the antenna is looking straight through the windshield.
There's a few drop outs, probably when I've moved my hand off the wheel. Confirms what I found with the two of them on the roof, the Mazda two speedo reads about 8% high.
So anyway, I've got 6hrs 49mins worth of data before the battery cut out.

But, the GW52 battery ran out during a distance attempt on 7/12/16. It went from 0858 to 1726, I make that 8hrs 28mins, so the GW52 has an extra hr and a half of battery life at the moment, some batteries are supposed to improve with a few charge cycles, I'll do another test in a week or so and see if it's changed.

Just been looking for specs on the locosys site and I can only find info on the GW52 battery life.

Select to expand quote

Operation time: up to 2 years for watch, 14 hours for GPS

So I only got just over half that the other day.

I'll hunt around a bit more see if I can find what they say for the GW60.

Well all I can find is the battery size which is 250mAh and the memory spec,

Select to expand quote

Built in flash memory to store 1,000,000 GPS logged points

If my maths is correct, that's 55 and a half hours at 5hz. If I'm right that's a big improvement on the GW52.
Yep sure is, GW52 is

Select to expand quote

Built in flash memory to store 12,000 GPS data

So next battery test I'll do at 5hz, because it isn't going to run out of memory, which is why I used 1hz in this test.