Sorry, Vosadrian, I just assumed the files were randomly jumbled, but it looks as though they came through in order.

I'm no longer on the advisory group, so my interest is only personal, others will do the real evaluation.

But a quick look, there is good conformity with the motion, (the OAO files are what's wanted, GPX doesn't include some of the data that's needed).

The positional data also looks OK, that's the dotted line, red is the watch blue the motion.

Could be I just selected a good bit, others will be more thorougher.

Thanks for looking at it decrepit. That was on a bike through Sydney hills, so mprobably some obstructed sky. I expect the data to improve on water with full unobstructed sky.

Probably opening a big can of worms but since a few have asked about Apple Watch on this thread and in the spirit of inclusivity thought I would share my findings after owning an Apple Watch Ultra 2 for the last couple of months.

In summary, in my understanding of this thread, results within +/- 0.2 kts are acceptable for GPSTC. Over the last 3 months I've done 2320km during 75 windsurf sessions in various conditions and locations. Data below. With the exception of 2sec data the rest of the categories from Apple Watch match very well, "almost" always <0.1kt, errors bigger than that are typically a penalty for the Apple Watch, or they are non competitive outliers..i.e. a NM with 4 gybes. The 2 sec data is as expected very unreliable because of 1 sec sampling with typical errors +/-0.7-1kt.

I purchased the watch 100% for communication and health aspects, not for any speed recording. I use 2 approved boom mounted units to log windsurf sessions for GPSTC. When I saw this thread I thought it would be interesting exercise to compare the new watch data with boom mounted units.

First point worth noting is that unlike Michael George who used the Waterspeed app to record sessions, I use the native Apple Activity app and record the session as a cycle ride. After the session the watch ports it to the iPhone and then I use the Healthfit app to auto sync it to Strava where you can download the GPS file. You can record session as a walk or run but then Strava flags it as dodgy due to the speeds so the cycle ride seems to be the best option to record natively on Apple Watch until such time Apple has a native windsurf option.
I haven't seen the watch lagging with initial acceleration like Michael did so suspect the 3rd party Waterspeed app is not getting the data from Apple Watch efficiently? The data seems to track very well with the boom units with no lag. In gybes the sail flip/hand movement nearly always generate something like the below comparison. Comparing two boom units shows similar and been discussed on the DIY GPS thread/YT so seems it is a reasonable artefact of hand movement that doesn't affect #'s in anyway.




For comparison I used GPSSpeed reader with default settings. I've only used one boom unit for the data comparison (as I only have historical files for one boom unit since this thread started) and not the average of 2 boom units. The boom units are always within +/-0.1kt for 2sec and 0.02kt for other categories.

The 2 sec as mentioned is problematic, not as bad as previous Apple Watches though. (Positive difference means Apple Watch reading higher than Boom unit). If you subtract 1kt in all cases you should have honest speed. The observation for me is multiple walking speed tests (not shown here) show +/- 1 kt difference in 2sec or nearly 40% error but above 30kts the error reduces to +/-0.4kts or around 1%....makes total sense for positional only speed...putting it out there before doing any comparisons, for the top 50kt sailors, the error may??? reduce to acceptable limits. The rest of us mortals it doesn't matter.





The 5x10 is much better with the results usually penalising the watch. There are three outlier sessions which are above +0.2kts and two of those sessions show glitches on the Apple Watch that I can't explain. The glitches seem to occur randomly and don't seem to be in the 5x10 calcs.





these the weird random glitches for the outliers for 5x10 but no glitches occur in the calcs..I've found glitches like this in 3 of the 75 sessions. They always have a spike then flatline. No clue at this stage why.




Alpha as expected almost always penalises the watch.




NM always very good with the outliers +/- being non competitive like having 4 gybes




Distance and hr are always slightly +ve due to the 1 sec location only sampling. The distance displayed on the watch in realtime is always different to what is ported to the iPhone. Seems there is some fancy filtering based on accelerometer data etc but in general distance recorded is close but always above. The crazy +ve outlier distance data is when I'm walking about (especially drone sessions) and the -ve data when Apple Watch went flat as I forgot to charge it sufficiently for session.









In conclusion I was quite surprised how accurate the results were and my take is other than the 2sec and a couple of undiagnosed glitches, results from Apple Watch Ultra 2 are on par with the Garmin's discussed here. Seems the data is a solid backup if nothing else. If Apple just did higher sampling might be another option for approved unit. For the masses though the results are pretty good for weekend warrior/no one really cares comparisons. All the raw data shown here publicly available on KA72/Strava or can send if anyone wants to check my results.

Select to expand quote

.........results from Apple Watch Ultra 2 are on par with the Garmin's discussed here.....

No, they are not close to being on par.
I have produced a graph of 80 sessions of my Garmin 255 vs a Mini Motion. I have set the scales to match your graph above.

Select to expand quote

tbwonder said..

No, they are not close to being on par.
I have produced a graph of 80 sessions of my Garmin 255 vs a Mini Motion. I have set the scales to match your graph above.

You right TB wonder, the 2sec from Apple Watch aren't accurate and will never be due to the 1 sec sampling. Whilst the 2 sec is not on par, the error seems to be almost always less than a knot. So you could just subtract 1kt at all times to have a reasonable idea without being unfair to other sailors.The point I was trying to make is the other categories appear 'on par' with the Garmin with a roughly +/- 0.1kt difference (bigger errors as mentioned are either lower or non competitive). Since the consensus seems to be a 0.2kt error is acceptable for GPSTC and most recreational sailors it seems the data should be at least considered to widen the appeal of GPSTC to a wider user base. The Ultra 3 is rumoured coming around Sept '25 is supposed to have blood pressure, satellite comms and hopefully better GPS. I for one will trade my current watch in for that and do another data comparison.

Nice analysis Flex2, thanks for taking the time to share it.

Thanks for sharing Flex2.

A friend sent the Apple Watch Ultra data to me and asked what I thought. Rather than respond to one person, I uploaded those findings but I've not really looked any more closely because COROS and Garmin have taken up all of my spare time.

The OS provides location data (including speed, etc) via a common class, so it's unlikely that Waterspeed is breaking things imho. I'd also imagine that Apple implement bespoke sport profiles in much the same way as Garmin, COROS and other watches.

With regards 2s speeds, your stats show what would be considered outliers in the Garmin data are occurring regularly on the Ultra 2. There are also some differences which far exceed anything that I've observed on my FR 255 or Fenix 7 Pro.

Can you perhaps upload your data to Google Drive (or similar) so that anyone curious can have a closer look?

I used my brand new Garmin 255 music today with the GPS Approved profile loaded. I'm really happy with it. Thanks TBWonder for your efforts

K888, trying to upload all the data but something funky going on with google drive today..can't upload a single byte.. will try tomorrow

Select to expand quote

Flex2 said..
K888, trying to upload all the data but something funky going on with google drive today..can't upload a single byte.. will try tomorrow

Thanks, much appreciated.

ok, here is the link to the files and spreadsheet organised by date. Note the earlier files only have one boom unit file, then a few have both boom units, then I grew lazy and only downloaded 1 as the boom units always match extremely closely. Jan 6 is the one that has the funky glitches on the Ultra data and the worst 5x10sec
drive.google.com/drive/folders/1XEKgqQaQqEhJ9wu0uCtQo9S9LR1HEdKI?usp=sharing

To all working on this, know it is appreciated.
As an intermediate, i love recording data and reading peoples sessions.
Fractions of a knot hardly matter to me at this stage, but i see how it does to the experts and record setters.
I think there is more intermediate sailors out there who'd take part using watches if they can be accurate, and i see a bright future for windsurfing in recording and comparing GPS data within an online community. Hopefully your efforts will bring more sailors into it and grow the fun.

Flex, thanks for sharing results and data. Given that the Apple watch speeds are position based, the accuracy of the results is quite good. That said, the observed errors are quite a bit larger than seen for the Garmin 255; in my opinion, they are not good enough for GPSTC, since they would affect the ranking.

Unfortunately, your work flow resulted in GPX files that do not include doppler speeds. The cause of the larger errors is not the 1-second sampling, it's the use of positional data. It would be quite interesting to see what the doppler speed results would look like for your sessions. I don't know if the Apple Activity app records doppler speeds, or if you still have the original files if it does.

Boardsurfr, no Apple Watches currently record doppler speed so the GPX files is it as far as I am aware. You right, I don't think it's good enough for GPSTC but for the average sailor the results (other than 2 sec) are good and consistent enough that you have fair idea where you stand. i.e. if you are getting 22+kt Alphas from the Apple Watch Ultra but aren't currently in GPSTC its probably worth getting a approved device and joining a team.

Select to expand quote

Flex2 said..
Boardsurfr, no Apple Watches currently record doppler speed so the GPX files is it as far as I am aware.

That seems to disagree with what Mike wrote in the various pages at logiqx.github.io/gps-details/devices/apple/watches/sessions.html

Watches that report speed as a separate data field usually use doppler speed. But unfortunately, speed is not defined in the official gpx definition (only in various extensions), so speed is often skipped and lost when exporting to gpx.

Select to expand quote

boardsurfr said..

Flex2 said..
Boardsurfr, no Apple Watches currently record doppler speed so the GPX files is it as far as I am aware.

That seems to disagree with what Mike wrote in the various pages at logiqx.github.io/gps-details/devices/apple/watches/sessions.html

Watches that report speed as a separate data field usually use doppler speed. But unfortunately, speed is not defined in the official gpx definition (only in various extensions), so speed is often skipped and lost when exporting to gpx.

ISTR the Healthfit app (used by Flex) is one way may save the Doppler speed data to GPX files, but I may have mis-remembered. I'd quite like to see those files, if Google Drive is playing ball now. In the absence of that, Flex could share a single GPX with either of us to confirm.

The Waterspeed app captures the speed data from the location API and includes it in their GPX exports. They have a bunch of different GPX export options (although none adhere to XSDs), but they do at least include the Doppler speeds.

would be great if Apple used/recorded doppler speed but 100% of all my research say they don't. Hope to be proven wrong

Select to expand quote

ISTR the Healthfit app (used by Flex) is one way may save the Doppler speed data to GPX files, but I may have mis-remembered. I'd quite like to see those files, if Google Drive is playing ball now. In the absence of that, Flex could share a single GPX with either of us to confirm.

hmm, I posted link to all files this morning..
here again if you missed it
drive.google.com/drive/folders/1XEKgqQaQqEhJ9wu0uCtQo9S9LR1HEdKI?usp=sharing

Select to expand quote

Flex2 said..

ISTR the Healthfit app (used by Flex) is one way may save the Doppler speed data to GPX files, but I may have mis-remembered. I'd quite like to see those files, if Google Drive is playing ball now. In the absence of that, Flex could share a single GPX with either of us to confirm.

hmm, I posted link to all files this morning..
here again if you missed it
drive.google.com/drive/folders/1XEKgqQaQqEhJ9wu0uCtQo9S9LR1HEdKI?usp=sharing

Thanks.

Strava drops the speed data which is why it's not in those GPX files. Can you download FIT files from the Healthfit app? Those might contain the actual speed data, but I don't know about their GPX export (maybe not).

The majority of my Apple analysis (Apple Watch 8 and SE) was based on Waterspeed and Hoolan data which definitely contained Doppler-derived speeds. I was assisting the author of Hoolan last year and his app definitely records Doppler-derived speeds.

Well, something new..ya can download the FIT files and import to GPSSpeed reader and first pass look they don't flag as positional data like the GPX data does...but GPSSpeedreader reports speed as positional speed anyway, but calcs are different to the GPX files.. No doubt Boardsurfr can make more sense of this. Have uploaded just Jan 30th FIT file to google drive as of now. Will upload more FIT files soon

Have uploaded FIT files Jan 24-30 and the funky Jan 6 data to google drive for those interested. Can port more if anyone sees a benefit to do so.

Thanks Flex.

Viewing one of the sessions in Speedreader, it is clear that the postitional data (blue) is being quite heavily filtered. To be fair they seem to be doing a reasonable job of that but it is still positional data.

What should be the Doppler-derived speed (green) is also fairly decent at times, but also quite poor at other times. It's certainly not looking to be on-par with the Garmin watches, so probably has some way to go yet.

It's how faithfully a device captures the peaks and troughs that differentiates good from bad performance imho. The peaks affect the 2s + 10s speeds (fairly decent performance for much of the FIT), whereas the troughs affect alphas, NM, and gybe statistics (not great).

The FIT is certainly more appropriate than the GPX based on this very quick check.

There may be a more appropriate activity profile than cycling, ideally bypassing the Apple filtering. Anything expecting arm movements won't be any good (running, etc), but perhaps there is a generic profile like on the Garmin watches?

p.s. I'm not sure the Garmin thread is the correct place for this discussion?

Select to expand quote

Flex2 said..
but GPSSpeedreader reports speed as positional speed anyway, but calcs are different to the GPX files..

Not sure why you are saying this. The current version of GPSSpeedreader issues warnings about GNSS settings ("not all GNSS + multiband"). Older versions warn about missing satellite data and accuracy estimates. Maybe you are using a really ancient version? If so, current versions are at github.com/prichterich/GPS-Speedreader

Select to expand quote

K888 said..
Thanks Flex.

Viewing one of the sessions in Speedreader, it is clear that the postitional data (blue) is being quite heavily filtered. To be fair they seem to be doing a reasonable job of that but it is still positional data.

What should be the Doppler-derived speed (green) is also fairly decent at times, but also quite poor at other times. It's certainly not looking to be on-par with the Garmin watches, so probably has some way to go yet.

It's how faithfully a device captures the peaks and troughs that differentiates good from bad performance imho. The peaks affect the 2s + 10s speeds (fairly decent performance for much of the FIT), whereas the troughs affect alphas, NM, and gybe statistics (not great).

The FIT is certainly more appropriate than the GPX based on this very quick check.

There may be a more appropriate activity profile than cycling, ideally bypassing the Apple filtering. Anything expecting arm movements won't be any good (running, etc), but perhaps there is a generic profile like on the Garmin watches?

While visual comparisons can be useful, I'd be careful about the conclusions - in particular about the "not great" about troughs. The troughs you show are all in jibes with a late sail flip. One unit is on the boom, the other on the hand, which travel quite differently during the sail flip. Given that the Apple watch doppler speeds otherwise match the u-blox boom GPS data quite closely, it's quite possible that the differences are real. Here's an example from one of Jim's sessions showing big differences between the two boom units in a jibe:

Several points before the selected point show a speed difference of 2-4 knots. That's a true difference, from one side of the boom making a different circle than the other one. You can see something similar in many jibes when using 2 boom units. The hand that the watch is on will move different from both boom units, so show different speed. The timing will also be a bit different, with more hand movement after the sail rotation is finished. When the hand moves back on the boom, that's in the opposite direction of travel, so it will reduce the measured speed.

But it definitely looks like the positional data are heavily filtered, while the doppler speed data are not. It is very possible that the observed differences will be lower for the doppler speeds. But if you want to compare 1 hour data or distance to the boom units, you need to consider any breaks taken, where the sail was stationary while the watch recorded speeds while walking around. Either use a relatively high minimum speed filter (like 5 knots), or erase points in break sections.

Apple musings resume at www.seabreeze.com.au/forums/Windsurfing/Gps/Apple-Watch-Ultra2?page=1#5