Garmin 920XT Windsurfing activity

Even more confusing Ricks .gpx file definitely has no doppler data in it but Fin's from 22/10/15 does. It has several 127Kt spikes that get past GPSResults's filters. After removing the spikes it has a max difference to the GT31 2s results of 0.763kts.
I've yet to look at the rest of Fin's files, or work out if the 127kt spikes coincide with any particular tack. Rick's file has a slightly more spiky nature on one tack than the other. Pointing to a possible underhand grip problem.

Select to expand quote

decrepit said..
Even more confusing Ricks .gpx file definitely has no doppler data in it but Fin's from 22/10/15 does. It has several 127Kt spikes that get past GPSResults's filters. After removing the spikes it has a max difference to the GT31 2s results of 0.763kts.
I've yet to look at the rest of Fin's files, or work out if the 127kt spikes coincide with any particular tack. Rick's file has a slightly more spiky nature on one tack than the other. Pointing to a possible underhand grip problem.

??? Any device that can put a random 100kt spike in it, needs to be seriously questioned as to its validity. ie: Underhanded or not, the GPS-engine shouldn't randomly just generate bad data -> it should be "well behaved", such as simply writing zero's or leaving that timestamp empty.

Select to expand quote

decrepit said...
Now what idiot has red thumbed Sailquick's above post????
It's stating that the GT31 is still the gold standard for econmical, verifiable accuracy.
Sure there are cheaper units, there are units more user friendly, but none of them store the info needed to verify the data. If this isn't important to you, then you don't come into the catergory that are serious about rankings.

I keep on using the Garmin 'something' as long as it can give me numbers and am not fuzzy about being that accurate but in the world of GPS rankings here and overseas what are the acceptable brand and models so we can keep abreast?

Select to expand quote

sick_em_rex said..
did you happen to ask Rick if he sails with an underhand grip? From when I have sailed with him he doesn't, so I don't think that is the issue.
Also many SUP riders are using this device and their hands are all over the shop changing from side to side and they don't seem to suffer spikes like this. Is there a chance it could be your software Mike?

The use in running sports and Paddling sports throws up a very different set of issues for GPS design.

Imagine the acceleration back and forth from a runners arm or the rhythmic back and forth of the paddlers arm. This should produce Doppler speed data that undulates sharply up and down on a speed graph. The Doppler speed reading (and 1hz positional for that matter) would be very inaccurate and almost certainly useless for any kind of competition. It would have massive aliasing issues and the speeds recorded at 1hz using Doppler would be all over the place. I initially speculated that they put clever filters in the GPS engine to 'smooth' this out, but I can't see any good way to do that. I think they use the trackpoint (locational) data for this reason. If you only take a reading every few seconds, or better, tens of seconds, you would get a rough estimate of your average ground speed. Close enough for runners and paddlers who don't use the data for any kind of competition. For their comps, they always have a start and finish line. Even if they use a GPS coordinate for that (which would surprise me anyhow) the Trackpoint method would work 'OK' for distance between two points a long way apart.

Looking at Ricks tracks it is pretty clear that is more 'noisy' speed data on alternate tacks. It may not be underhand grip, although we think it is consistent with this. It may be just back hand, front hand difference in positioning and reception, or maybe something else....

Anyhow, the point is that fitness devices and lots of other GPS based devices don't have anywhere near the pure speed accuracy requirement that we have. They just don't need it, and in the way they use them, they will likely never achieve it without for, example, next generation RTK type accuracy at very high HZ where they may be able to 'filter' the data with clever algorithms to 'smooth' it into something usable. And even then, such filters would likely play havoc with our requirements for unfiltered data.

The GPS 'sports' watches are very, very good at what they are designed for. But they are not designed for our requirements which are very different.

Fun session today. Once again, the Garmin comes in a touch slower than the GT31 and a bit shorter in distance.

GT31
2 Second: 29.44
5x10: 27.577
1 Hr: 15.888
Alpha: 21.267
Nm: 20.706
100m: 29.054
Klms: 37.828

Garmin F3
2 Second: 29.365
5x10: 27.571
1 Hr: 15.845
Alpha: 20.998
Nm: 20.712
100m: 29.003
Klms: 37.377

On the fr920xt and probably also Fenix3 make sure to base your new activity profile on "Other" and not "Running" or so. That way you get much better recording. I think they have some special algorithm for Running.

Here would be some step by step instructions: chumba.ch/garmin/instructions.html

Today I might get a new session recorded with the Canmore, fr920xt and GT-31. Let's see. It's on ice :).

Three GT 31s same software and settings

Uploaded with GPSAR Pro

GT 31 (1)
32.20 30.11 14.35 19.96 24.02 51.746

GT 31 (2)
32.24 30.07 14.35 19.96 24.01 51.763

GT 31 (3)
32.29 30.08 14.34 19.96 24.02 51.661


Uploaded with KA 72

GT 31 (1)
32.054 29.914 14.843 19.664 23.719 51.15

GT 31 (2)
32.122 29.927 14.845 19.689 23.762 51.169

GT 31 (3)
32.132 29.929 14.844 19.694 23.764 51.136

Select to expand quote

AUS4 said..
Three GT 31s same software and settings

Uploaded with GPSAR Pro

GT 31 (1)
32.20 30.11 14.35 19.96 24.02 51.746

GT 31 (2)
32.24 30.07 14.35 19.96 24.01 51.763

GT 31 (3)
32.29 30.08 14.34 19.96 24.02 51.661


Uploaded with KA 72

GT 31 (1)
32.054 29.914 14.843 19.664 23.719 51.15

GT 31 (2)
32.122 29.927 14.845 19.689 23.762 51.169

GT 31 (3)
32.132 29.929 14.844 19.694 23.764 51.136

so what is your point of posting this? ... as it appears that you have your GPSAR Pro setting, set to something different that the non-configurable setting used on KA72.

Or is you point that 3 non-clock-locked GPS's are producing 3 slightly different results?

Or some other point?

Select to expand quote

N1GEL said..
Fun session today. Once again, the Garmin comes in a touch slower than the GT31 and a bit shorter in distance.

GT31
2 Second: 29.44
5x10: 27.577
1 Hr: 15.888
Alpha: 21.267
Nm: 20.706
100m: 29.054
Klms: 37.828

Garmin F3
2 Second: 29.365
5x10: 27.571
1 Hr: 15.845
Alpha: 20.998
Nm: 20.712
100m: 29.003
Klms: 37.377

Hi Nigel. Have you sent the files to Mike to add to our collection of statistics? If not, please do.

Select to expand quote

sailquik said..

AUS4 said..
Three GT 31s same software and settings

Uploaded with GPSAR Pro

GT 31 (1)
32.20 30.11 14.35 19.96 24.02 51.746

GT 31 (2)
32.24 30.07 14.35 19.96 24.01 51.763

GT 31 (3)
32.29 30.08 14.34 19.96 24.02 51.661


Uploaded with KA 72

GT 31 (1)
32.054 29.914 14.843 19.664 23.719 51.15

GT 31 (2)
32.122 29.927 14.845 19.689 23.762 51.169

GT 31 (3)
32.132 29.929 14.844 19.694 23.764 51.136

I suspect the point you are trying to make is that different analysis software (the correct term is analysed or analysis, not 'uploaded'!) give different results.

It is not only the settings of the GT-31 that have to be identical, it is the settings in the analysis software that need to be the same.

In the examples you posted, obviously the settings in GPSAR-Pro are different from those in KA-72, and not set correctly. I regularly run files in all the alternative to check this very thing and do not see variations like that.

When set up correctly with the default settings, GPSAR-Pro, GPS-Results and RealSpeed get virtually identical results to KA-72.

There are a very few known issues with some types of files and GPS settings with some of the alternative softwares, but they don't affect 99% of people and files.

As far as I am aware, the various programs come set up correctly by default, or at least acceptably correct. The only way I can see for the above results is that someone has fiddled with the default settings in GPSAR-Pro.

Please send me the files and I will run them in the various analysis programs with default settings and post the results.

Nah, I was sailing/chatting with Rick yesterday and pretty sure his point is that 3 identical devices will still produce different results.

Andrew, just my personal preference on this. I prefer to do my own processing, because I then have a better chance of knowing what's going on.
Especially using GPSResults that gives a few different looks at the SDoP data. I can easily pick which GPS, worn where, gives the most accurate results.
I can also see exactly what the various filters are doing and why.
Rick's original watch file, is a case in point, it was full of holes, (because the watch had been set up wrong), and the KA72 analysis didn't filter out a big spike.
Without seeing the data, you don't know what condition it's in, or how credible the results are.

About the 2 missing runs on the fr920xt:

It might be that I paused the track and forgot to start it again. This sometimes happens to me, because I usually pause the track before a break to keep the average speed a bit higher. As you can see in the screenshot there was a break in both tracks, but then I forgot to start it again. The fr920xt continues right after 2 legs, when I noticed that it was not recording.

I wear the fr920xt on might left wrist and the canmore and gt-31 are in a waterproof bag around my neck, about 10cm below my cheek. I can do some more tracks with the canmore and gt-31 also in my left arm.

My front hand does not grab the boom from below, so the watch was always pointing upwards.

For what its worth, I use the same GPS (GT31) and analysing program (KA72) for all my sessions. I do this to have like for like comparisons between sessions.

Select to expand quote

sick_em_rex said..

Agreed, that is exactly why he was doing it.

Andrew, your points above are quite alarming as you yourself say that different programs will give varying results and even to the point that unless you have the program you use setup correctly then it also will give different results.

Correct. If the setup is wrong, the results will be wrong, or in this case may be different. Nothing unusual or new about that concept.

Fortunately, the programs come set up correctly by default. To get them to give erroneous results you would have to deliberately fiddle with the settings whilst not knowing what you were doing. Never a good idea in any software.

Select to expand quote

sick_em_rex said..This opens a whole Pandora's box as to what programs people use and the potential for huge variances.

Not if users don't go fiddling about with settings they don't understand, and the differences potentially created are not huge at all but actually very small

Select to expand quote

sick_em_rex said..This is regardless of what GPS device is being used (as evidenced by Rick's 3 identical devices ((GT31's)) using up to date firmware and getting different results for all 3).

See my previous post explaining that the differences are extremely small from identical side by side devices and easily explained.

Select to expand quote

sick_em_rex said..What is the tolerable variance you allow for? If looking at Rick's results, there is a potential for a tenth of a knot in 2 sec speeds which could be the difference between someone breaking the desirable 40 knot barrier and not (or simply gaining a PB). I have seen people miss 40 knots by less than a tenth of a knot and I know if it was me then I'd be pretty miffed if I knew that there is potentially a program someone else is using that may not have robbed me of that tenth. I have two identical (to 3 decimal points) alpha pb's...who knows, using different software I may have bettered it???.

Yes. See the aforementioned post regarding the lower accuracy of the 2 second calculation. Common potential maximum error for two second peak speeds in the GT-31 are in the order of up to 0.25 seconds. However, it is usually less than 0.05 seconds in the hundreds of tests I have done. The 2 second category is inherently the least accurate and should be not taken too seriously.

Select to expand quote

sick_em_rex said.. I am not a dumb person (open for debate however ) but I don't think I would know how to setup correct analysis software and I wouldn't know whether it was setup rightly or wrongly anyway. I plug my SD card into my puter and let KA72 do its thing. I think this is why so many people want to use a standard program to read all files. I don't care if it's KA72, Realspeed, GPSAR or GPS Results. Otherwise we are going to go round and round in circles and this debate will last for all of eternity.

Quite correct Andrew, you are not at all a dumb person. You are probably just asking the same questions that other smart people are thinking. It's all good.

As I have pointed out, if you use any of the analysis programs, you don't really have to know the hows and whys of the settings. They are mostly buried where normal people won't go digging anyhow and are set of correctly by default. There are various guides and discussions you can search for and study if you really want to know about those things, but it is all unnecessary for simple plug and play analysis. Putting your SD card in your computer and letting KA-72 do it's thing to letting GPS-Results or RealSpeed or GPSAR-Pro do it's thing should always get the same results to the significant decimal point. Thats my point. I have explained in some detail in another post why there is not such thing as the concept of a single 'standard' program. It's more like there is a set of definitions of categories and specifications of implementation to calculate those things. Each software engineers might go about achieving this in a slightly different way to get the same results from good clean files. We only see significant differences in the results when the files have some types of errors and because each program deals with them in different ways, we often have to try analysing in a number of different programs to find the source of the error and see if we can get a reasonable result for that file.
Every program, including KA-72 will get rubbish results from some types of rubbish files. There is no such thing as the perfect, 'standard', never erroneous analysis software. We will always have to view any result they produce with logical, considered awareness. If it looks wrong, it probably is wrong, and we need to do deeper analysis to see if we can get a true result from it, or just bite the bullet and throw it out and try again. (Not so bad, we get to sail again! )

Select to expand quote

sick_em_rex said..Just a question, what is it that makes your software analysis program any better than KA72? And please try and reply in layman's terms, my brain zones out when you get all technical

None of the software programs are 'mine'. But I use them all. Refer to my answer above. None is perfect, but all do a great job with good clean files from the approved GPS devices.

Further to last nights answer, I'll try and present some of the problems as I see it now. Doing these various tastings has given me a deeper insight into what's going on.

The GW52 5hz ripple is the main factor. I didn't think my speed could vary that quickly, but when I wore two of them on my head, in the same pouch, they both had the same waveform.



Bear in mind this is a 5hz ripple not 1hz.
So then I thought my quick speed changes where real, but then I think it was Matthew who pointed out that both units were subject to the same interference.
And this was confirmed a short time later when I had one unit on my head and one on my arm, this produced a different result. (I'm sure I saved a pic but can't find it now.)
But the 5hz ripple on the two wave forms was out of phase, instead of going up and down together as above, where one went up the other went down. This indicated to me that the ripple was real, but nothing to do with my speed! It had to be interference/noise of some sort. My guess is a radio signal of some sort, possibly a mobile tower near by.
The GT31 has a low pass filter that prevents this higher frequency ripple getting through, the GW52 just averages the ripple out.
We have no idea how any other device handles this, if the filtering isn't adequate then the 1sec reading can fall anywhere on the ripple, 2 consecutive highs will give an abnormally high 2second result, and conversely 2 points at the dips will give a low reading.
Just the odd comparison of one or two results won't necessarily show this up, that's why we look at multiple 2s results from the one file.