Alternative to the GT-31

I looked into the GP102 .fit files some more. It turned out to be rather easy with the FIT SDK that can be downloaded from www.thisisant.com/resources/fit. The SDK contains an executable jar file to create text files in .csv format from .fit files (FitCSVTool.jar in the java folder). Here's an example of what the result looks like:



So the files contain latitude and longitude data. They are encoded as 32-bit ints (which is what the "semicircles" means), which gives a granularity of about 1 cm, more than enough. Kind of cool to get the .csv file so easily, lets you play around with the data. For the 001.fit file in the upload I mentioned earlier, the median difference between positional speed and doppler speed is 0.59 km/h. The maximum is 44.2 km/h, which was a spike in the positional data after a crash where the GPS lost reception for about 10 seconds. Doppler speeds during this time were < 1 knot.

pheww; thanks Boardsurf, I just bought 3 canmores so happy to hear they are precise !!! (especially considering I'm starting a freerace challenge locally !!)

Select to expand quote

seanhogan said..
pheww; thanks Boardsurf, I just bought 3 canmores so happy to hear they are precise !!! (especially considering I'm starting a freerace challenge locally !!)

Where did you get them? Ive only asked a few times, Local shop or on line?

Select to expand quote

snides8 said..

Ha still got mine! Still works even saw a few 30s on it back in the day

I might have a couple tucked away in my windsurfing museum as well! Come to think of it, I should get one out and compare it with the GPS readout!

Select to expand quote

boardsurfr said..

No, we do not. You may want to read sailquik's original post a bit more carefully. What is needed is some kind of additional data besides the doppler speed that makes it possible to spot errors. Accuracy estimates like Hdop are great for that, assuming they are reasonably accurate. Information about the satellites tracked can also be useful (for GT-31 data, Hdop often seems to be almost perfectly correlated to the number of satellites tracked). But even error estimates like Hdop have their limitations, which is why record attempts require data from two GPS units.

Trackpoints, i.e. positional data, provide another option to check doppler-based speeds. Differences between positional speed and doppler speed can be used to identify questionable data points which then can be excluded. They also can be used to get a quick idea about data accuracy by simply comparing speed and doppler speed in tools like GPS Action Replay Pro. For all the Canmore GP102 data that I have looked at, those difference were very small (except after crashes when the unit was under water, and positional speeds sometimes showed spikes).

sailquik mentioned some of the very old units which provided only position-based speeds, and had very poor accuracy. He also had the wrong impression that the GP102 does not provide track points. If that was indeed so, that would mean there would be absolutely no way of knowing how accurate the data for any given run are. But it's not - the Canmore GP102 logs positional data as well as doppler speeds. So ..

sailquik said..
In that case the data from the Canmore just got significantly better than any other GPS except the GT-31 which also includes error data.

My statement of coming full circle must be lost in translation. This is me specifically saying, "the Canmore might be fine, but missing error data - since you spend a few thousand on gear, why dont you spend a few 10 more to purchase a gps which *does cut the mustard*..?"


For record attempts, have you ever seen the final-record showing the result of two gps' combined? You wont - the data for the record, doesn't use two GPS's results. That requirement is so that the first units' data, can be compared to the second unit -> to verify that they both look the same.


Given the way you describe "HDOP as being useful if reasonably accurate" indicates they you may not understand where it comes from [ but I apologize if I am mis-understanding ].... HDOP is output data from the maths that are being undertaken to calculate the position. In that vein, the information about how many satellites are tracked, is already being "used" (for want of a better word), in the HDOP... so the xDOP accuracy is *by definition* THE accuracy of a Geodetic model. So the additional data that we should be using, *is* the Doppler SDOP data (or whatever you want to call it).


Note that trackpoints dont verify Doppler data... they are about x100 less accurate.... that would be like using your car to measure the length of toes on your left foot. That is what the second GPS is for.


It doesn't matter what "Daffy thought the file-format had, so the Canmore must be accurate"... that is completely irrelevant -> either a) there is error-data which says how inaccurate your GPS is, or b) you dont know how accurate it is. ... Does the Canmore produce error-data ?

What we do know about the Canmore is, "the data correlates well with existing devices that have been well tested". AFAICR from various posts, the Canmore sometimes does show larger variation, that exceeds the variation between two GT-31's.


So if you care whether your "data can be verified be the geeks", then spend an extra $50 on a GT-31 and get the benefits that that extra $$ provides, such as the Speed-Genie.
Or dont spend that cash, but then try to claim it is equally accurate.

Select to expand quote

mathew said..
This is me specifically saying, "the Canmore might be fine, but missing error data - since you spend a few thousand on gear, why dont you spend a few 10 more to purchase a gps which *does cut the mustard*..?"

That's a pretty moronic statement, considering that the GT-31 has been unavailable in many parts of the world for almost a year now. The GT-31 was three times as expensive than the Canmore ($150 vs. $50); some of the replacements that have been discussed are 6 x as expensive. That may not matter to you, and to anyone who gas no problem spending thousands on a single rig. But many windsurfers I know are on a limited budget, replacing gear only when it breaks, and even then with cheap used gear. They may consider spending $50 on a cool gadget, but won't consider $150 or $300. That's even true for a bunch of guys who would have the money. Price is not a big issue for those already deep into speed surfing; it's a big issue when trying to build up speed surfing, as we do here.

Select to expand quote

mathew said..
Given the way you describe "HDOP as being useful if reasonably accurate" indicates they you may not understand where it comes from [ but I apologize if I am mis-understanding ].... HDOP is output data from the maths that are being undertaken to calculate the position.

Apology accepted. I happen to have a Ph.D. in experimental sciences, and have worked on error estimates for years. So I understand that error estimates are only as good as the assumptions they are based on, and the math used to calculate them. Looking at the numbers of xDOP from GT-31 data, it is pretty obvious that (a) the resolution of the accuracy parameter is quite limited, with +-20% changes being typical; and that (b) the numbers are very closely tied to the # of satellites tracked. I have not found a published description of the math used to calculate the xDOP parameters, but it seems very likely that it uses only satellite information (both number and position). It that is indeed true, then one can easily approximate the xDOP values from the number of satellites tracked.

Select to expand quote

mathew said..
Note that trackpoints dont verify Doppler data... they are about x100 less accurate....

Think about this for a minute and examine some actual data from a newer GPS unit, and you may just realize that you are wrong. The low precision that you are referring to is the absolute positional accuracy of a single data point. But for comparing speeds, we are only concerned about the relative accuracy of subsequent data points, which is about 2 orders of magnitudes higher. Decent newer GPS units like the Canmore GP102 or the Suunto watches have a lot of filters and advanced math that goes into the calculation of track points. If you look at Tom Chalko's SDOP paper that is the basis for all the desire for error estimates, you'll see that a lot of his conclusions are based on measuring errors on a stationary unit. If you look at the corresponding stationary data from a Canmore GP102, you can see that (a) the doppler accuracy is similar to the GT-31 accuracy, and (b) the positional speed accuracy is similar to, and quite possibly better than, the doppler speed accuracy.

Therefore, positional data can be used to identify problems in doppler speed data. You cannot simply use the same math to calculate a lower bound of the speed that was reached using the math derived for SDOP data, but that matters only for record attempts. For the 99+ percent of windsurfers and sessions that don't try to set official records, it does not matter.

Select to expand quote

mathew said..

It doesn't matter what "Daffy thought the file-format had

It matters quite a lot what Daffy thinks and says. He voiced a very negative statement about the Canmore GP102 that was based on some wrong assumptions. I simply pointed out that these assumptions were wrong, and provided the data to verify that. Daffy then revised his statement about the Canmore to a much more positive one. Nobody ever claimed the Canmore GP102 with it's current firmware was appropriate for record attempts. Almost nobody cares.

Select to expand quote

mathew said..

What we do know about the Canmore is, "the data correlates well with existing devices that have been well tested". AFAICR from various posts, the Canmore sometimes does show larger variation, that exceeds the variation between two GT-31's.

The only posts that I have seen which reported larger differences were talking about the "long distance" numbers - for example not being able to get a 1 hour reading from analysis software. That ended up to really be a math issue in the software used, which did not use total elapsed time - total distance. If someone reported data where the Canmore showed higher speed that two GT-31s, I'd appreciate a link.

Very interesting stuff boardsurfr. Thank you for your work with the Canmore and your excellent insights. As already stated, I agree that the availability of the trackpoint data means that the Doppler speed data from the Canmore can be at least compared. It does make spotting errors much more likely and makes the data from these units far better than the old trackpoint only Garmins etc. and probably on par with the GT-11 in that respect.

The GT-31 does not rely on HDOP to get speed error values. It uses 'SDOP' (Speed Dilution Of Precision) for accuracy monitoring. This is a term coined by Dr. Tom to for convenience and is not a standard industry term AFAIK. In the SIRF literature they refer to this type of calculation as Estimated Horizontal Velocity Error (EHVE).
Data of this type is also available in the UBlox .ubx binary output from their modules, but is not normally available from the SIRF chips unless you get the specific SirfDrive modules and the associated firmware. My understanding is that Locosys used the SIRF SDK to write their own firmware to get the parameters needed for SDOP specifically for the GT-31. Unfortunately, other GPS using the SIRF3 chip do not usually have this capability switched on.

Which brings us back to the Thingsee One. It uses a UBlox GPS that outputs .ubx binary files (at up to 10z) which include speed accuracy data. UBlox GPS have been tested specifically in speed windsurfing situations by a number of people and their abilities are fairly well known. I took part in one test which compared their results with the WSSRC fixed timing at Luderitz with excellent results.

The Canmore GP102+ output has Doppler speeds and more than decent accuracy, despite NOT being a dedicated GPSSS unit.
And, another thing. Considering Doppler speeds from a SIRF Star IV unit. May their accuracy be affected by lack of >1Hz logging frequency and xDOP; and do they depend on firmware?

Reliability is a disputable term. I understand if one wants all datas are OK without need to check, then he needs all that stuff, 10Hz, xDOP, ...; but if one is OK with checking the top results for no spikes, same as it has been so far, maybe he can do without.

0,1 knots accuracy "is still not good enough for meaningful competitions and rankings"? Really? Come on, do you really think one is faster or slower than someone else because of a <0,1 knots difference? Would you lightheartedly hold up a trophy won by resulting 0,05 knots faster than your buddy in #2?

I think when we say "feasible" we mean different things. Maybe you mean "technically" while I mean "economically"?

Select to expand quote

geoITA said..
I start to be confused: if the firmware can be customized (like in the GT-31 and - so they say - in the Walker), am I wrong or the limit is the chipset then? And, assuming one is happy with just Doppler speed and position data accurate enough to get speeds accurate to say +/- 0,1 knots, nothing else (no 10Hz, no DOP), would usual NMEA format work for him?

NMEA data does not necessarily get you reliable accuracy to 0.1 Knots at all! But even if it did, That is still not good enough for meaningful competitions and rankings

geoITA said..
No thanks, much smarter people weren't able to have one built or adapted to appropriate standards, so who am I to find that magic logger by myself. I will do with what I find, the Canmore will be OK for some time at least. Again, I am just suggesting those who can to ask for feasible things when they talk to manufacturers.

Which is exactly what we have been doing. I think we have just gone around in a circle here.

Apart from the fact that I never wrote "blah blah the canmore is OK for me" anywhere. At least because the Canmore has too small a screen to look at on the beach through a bag, and is not waterproof.

Now if we want to be serious, I think we should try to understand what the other is saying.

I am trying to bring it down to a few concepts in order to see whether "geeks" and "normal users" ("What is a 'normal' sailor? One who is interested in accuracy, not fantasy?", remember?) may be OK with the same kind of units, or if their needs are too much different.

Your statement about speed differences in rankings explains your point of view. What this means is, a speed difference of 0,01 knots between two guys that might have happened to do such speeds in different weather and sailing conditions and in places maybe thousands miles away from each other is something worth to be considered. And even need to know how much 'reliable' such difference is.

I used the term "feasible" a few posts above, as you answered to that I thought you knew that.

There are already a few very simple questions you could quickly and easily answer here if you wished to, should you ever find the enthusiasm. So I am starting to think that it's not the enthusiasm but rather the answer. An answer that fits with the general meaning of what you are saying, at least. Anyhow I understand you completely: if I was going round in circles, I would be done too.

And, I still have lots of fun sailing even when choosing not to use any logger.

Sorry for disturbing.

Select to expand quote

sailquik said..

geoITA said..
....Blah, blah, "canmore is OK for me..."

I agree and I am happy for you, shesh, whats your problem!??

geoITA said..
0,1 knots accuracy "is still not good enough for meaningful competitions and rankings"? Really? Come on.....

Definitely. Ok, here is just one example:

In the 10 second GPS-SS rankings there is 0.1 knot separating 2nd, 3rd and 4th. There is 0.01 knots separation 5th and 6th. Go on down the list and you will find example after example of less than 0.1kts separation groups of two, there and four or more sailors. If the accuracy is less than 0.1kts, these are virtually random placings. That is not nearly good enough in a high stakes competition!

It may be ok in the local lake league where no one knows any better, don't really care and there is very little riding on the results. Sure, no problems then. It may be OK for a 'fun ranking', but even 'fun' rankings have a way of getting pretty serious for some people.

geoITA said..
I think when we say "feasible" we mean different things. Maybe you mean "technically" while I mean "economically"?

I don't see where I used 'feasible so I don't quite know the context, but very large numbers of sailors spend large amounts of money on the latest and greatest gear, often with the idea that it will gain them some small advantage in speed over their friends or competitors. Put the price of a verifiably accurate GPS against that perspective.

geoITA said..
I start to be confused: if the firmware can be customized (like in the GT-31 and - so they say - in the Walker), am I wrong or the limit is the chipset then? And, assuming one is happy with just Doppler speed and position data accurate enough to get speeds accurate to say +/- 0,1 knots, nothing else (no 10Hz, no DOP), would usual NMEA format work for him?

Yes, I can clearly see you are confused. If you really want me to explain it to you, I might respond to a PM....if I can find the enthusiasm. I am done with going round in circles here.

I wish you great fun sailing with whatever GPS you choose to use.

It should work fine for you until/unless you find the need for something better.

Select to expand quote

sailquik said..
It should work fine for you until/unless you find the need for something better.

Yup. I got my first GPS logger, a GP102. I use one of those tiny Sistema food containers and a couple of ziplock bags. I haven't had any water inside the container yet. Maybe I'll try Vaseline over the o-ring. I'd be stoked for completely waterproof unit though...

It's really just a toy for now. It's a fun element to add to windsurfing for less than $100 dollars. If I can't put up 35 (or 40 or 45) yet, there's no point worrying about accuracy stats. GP102 and ka72.com is all I need for now.

Select to expand quote

boardsurfr said..

As much as I love the idea of SDOP, it's a bit frustrating. The SDOP values are not written to GPX files, and completely ignored by GPSBabel when it reads SBN files. GPS Action Replay appears to ignore them, too. It's almost impossible to find any information about it. Tom Chalko's original paper was great, but largely based on data points from stationary units. More troublesome, though, is that manufacturers did not even agree on how to calculate HDOP - seems that some used 50% values, others 1 or 2 sigma values or 95% values.

The Thingsee One sure seems exciting. I was pretty impressed with the ublox chip in the Flysight; it's SDOP was about 4x lower than the the GT-31 (and that was with an older chip). Add the higher data rate, and there is definitely hope for a lot of improvement in precision. I love the programmability of the Thing. Too bad they had a bit of a delay. But it seems we may have three good GPS units to choose from before the end of the year - a small, cheap one (the Canmore); a updated replacement of the GT-31 from Locosys; and a fancy Thing with better accuracy, resolution, and features for all those who don't mind spending a bit more and wearing a slightly larger unit.

I did not know that GPX files won't retain the SDOP values. I'm curious. Why do you use .GPX files? And what do you use GPSbabel for?
I also did not realise Yann has not implemented the SDOP values in GPSAR-Pro either. Had to check it again myself to see it. I admit I rarely use GPSAR-Pro now unless I am looking for something specific. I find it very unintuitive.

RealSpeed reports SDOP values, but Mal stopped developing it before integration of SDOP was fully integrated in the way he had planned.

Tom did quite a bit of mobile testing as well after he wrote the SDOP paper but didn't add it to his paper. I think he was satisfied the results backed the stationary data. He told me that he did a lot of mobile testing with the GX-52 prototypes and reported accuracy "10 times better" with that system, but despite my encouragement, he has declined to put his data on paper because he says he can't muster the enthusiasm. (He feels his efforts are not appreciated by the GPS speedsailing community, although, nothing could be further from the truth). He just says average accuracy is around 0.02 knots. But without data to back that up it's a pretty useless claim.

Locosys/Tom is very secretive about the method of calculating SDOP in the GX-52. They quote commercial secrecy. To me that is not good enough and very frustrating. To get any usable accuracy information on the GX-52 one would have to run their own tests. Again, not good enough!

At least with the UBlox chips, a lot of really good evidence has already been accumulated.

In my view the point is not about how expensive or not a proper gps should be, but rather how much meaningful speed differences of <0.1 knots are. Expecially in a sport like our and even more so when comparing speeds done in different days, different conditions and different places.

GeoITA. Where did you get the idea that the accuracy is as good as 0.1kts? Read my previous comment about this again.

Select to expand quote

sailquik said..
I did not know that GPX files won't retain the SDOP values. I'm curious.

GPX is an XML file format... it is extensible.

The term "SDOP" didn't exist prior to Tom stating it... prior to that the term was manufacturer (and literature) specific.
Maybe a quick email to the GPSBabel folks with a "please add <SDOP> as a value"...!