I have noticed that the Doppler speed in jibes is almost always a lot lower than the non-doppler speed (using a GT-31 and GPS Action Replay). The tighter the jibe, the bigger the difference. Minimum jibes speeds are often 1/3 slower with Doppler, sometimes more than 50% slower. That's true even with good GPS reception (8 satellites) and no obvious outliers in the positional track points.
The Doppler speeds also seem time-shifted relative to the positional speeds by 2-3 seconds (minimum Doppler speed is later). Anyone have a clue why?
The 1 Hz rate on the GT-31 is a bit slow for jibe analysis, anyway. Has anyone looked at the accuracy of newer 10 Hz units recently? It seems that the MTK chipset used in some of these can save Doppler data.
The inaccuracies are well known. Alphas are an informal category, not used for world records, which were pretty much invented to bring some variety to what was looking to be a discipline composed mainly of sailing in straight lines.
The most important factor, in my opinion, is that everyone's Alphas are calculated the same way. That minimises the effect of the inaccuracies by ensuring a level playing field, and allows for fair comparison of sailors.
(Unfortunately, there is no published standard method for calculating an Alpha speed, leading all the software designers to come up with their own method. Not that this has any bearing on your question, which goes more to device accuracy.)
Since we moved to the GT-31 from the GT-11 we have known that the Doppler Alpha calculations have shortcomings. This is the reason why the WGPSSRC has not created an official record catagory for Alphas.
However, as Dylan says, tests have shown that the Doppler data is still more accurate and reliable than trackpoint data. The biggest problem with trackpoint data is the false readings and spikes often clearly seen in the speed graphs during the sail flip part of the gybe on a large percentage of gybes. (This can happen with the Doppler data as well but is much less common)
10hz GPS with Doppler recording or transmitting capability are around in a couple of devices, but none that I know of have a screen readout. One I have tested that can record data can only record at 1 hz unless it is paired with a Bluetooth recording device. A few people have created such devices and have recorded 10hz binary data that contains all the info to calculate error margins. Those devices are based around a Ublox chip and output UBX raw binary data. Such devices can be approved for WGPSSRC records.
We have experimented with 5 hz Doppler data from a prototype GT-35. This had an MTK 5hz chip in addition to its standard Sirf3 chip. It recorded 1 hz data to the internal memory chip and 5 hz NMEA data to the SD card. It showed very good results for Alphas but unfortunately the MTK chips can't output raw binary data so they were not up to the job for records. A few hand made prototypes were made and tested but the project was not continued.
The ideal solution with currently known technology would be a device with all the features of the GT-31 but fitted with a Ublox6 10 hz GPS chip. Unfortunately, we have not been able to convince Locosys to build it yet.
Thanks for the info. I now understand that official records need raw data, and why there's no official alpha record. But for all the GPS users that are not in the running for official records (> 99%, I'd guess), the MTK chipset would be fine? I just found a waterproof 10 Hz logger that looks decent with the MTK set.
Regarding getting a better GPS for speedsurfing, I'd say think beyond the GT-31. Roo had pointed out a while ago that using smart phones and apps like GPS Speed Talker is the way to go. I tested that and found the smart phone accuracy lacking a bit (no doppler), but ended up improving my speeds a lot from hearing the actual speeds announced. So at least Bluetooth link out for Doppler data would be something worth adding. To bad that updating the GT-31 firmware effectively kills the bluetooth in the BGT-31.
Updating a device like the GT-31 that has very few users is a probably a money-loosing proposition for the manufacturer. There are a number of interesting-looking devices out there that aim at a much larger market. For example, the Bad Elf Pro has a GT-31 like display, 10 Hz logging, and bluetooth for interaction with GPS Speed Talker. They don't say much about logging formats or even # of data points, though.
If there is another file format that supports doppler (like SBN or SBP) I am always happy to extend support for it on ka72.com
It is not too hard to add additional file formats. Just a matter of picking apart the specs, or examining the open sources available.
Cheers,
Dylan.
The audio feed back in the GT-31 Genie can work the same way to help you improve your speeds.
There are some of us actively looking beyond the GT-31 for better solutions.
Can you tell me the make and model of the 10hz logger you mention. I have not seen one that actually logs internally at 10hz. Seen a few bluetooth dongles that can be used but they need a paired device to record to.
Andrew,
I just googled it and got here,
bad-elf.com/products/be-gps-2200
This is the interesting bit.
The 16 h battery life for the Bad Elf Pro is when connected with Bluetooth. Without, you get 32 hours. I contacted their support and got an response quickly - apparently, it supports 10 Hz logging, but there is currently no way to get Doppler data from files. Logging with 10 Hz would be limited to 10 hours (or perhaps less) due to internal memory limitations.
Another device that looks promising is from what appears to be a small hobby shop, the "SD GPS Data Logger V3.1.5" from Ohorarp LLC (http://ohararp.com/products.html). MicroSD card, 10 Hz logging with doppler, waterproof case, $125. Based on the MTK chipset, though, and no display. I contacted them, too, to see if I could buy a unit before an upcoming trip this week, but have not received a response in 2 days.
I have a post about GPS jibe analysis on my blog at boardsurfr.blogspot.com/2013/04/gps-jibe-analysis.html
Judging from the sample price, I can see why Locosys isn't too keen on creating a U-blox LEA-6T-0 based GT-31... ouch! The module does look very promising though, so let's hope prices will drop in the near future. Too bad none of the MTK modules are a go, they are much cheaper.
Andrew, one item you may have forgot about... the MTK chip, while at 5Hz, would average the values over a full 1 second. This is why I never bought a unit.
but I think Kato is 'taking the p1ss' (making a joke) about battery life since he's been known to sail for longer periods in one session (last time he did 540km).
I thought Kato did the record in just 14 hours - or am I confusing him with Jacques?
I downloaded the NMEA example file from the link at http://ohararp.com/products.html and looked at it briefly. There are discrete points every 0.1-0.2 seconds. The positional data look a bit funny, perhaps from lack of precision / too few digits - but the example is very low speed. The Doppler data look promising I think. I don't see evidence of averaging over 1 second, maybe it's a newer chip.
Boardsurfr. Kato did the most recent record in less than 16 hours but his previous record track had over 24 hours of continuous recording.
Tests done with 5hz and 10hz GPS in the past showed so much 'sawtooth' in the trackpoint/positional speed graph that the results were useless. Positional accuracy is just not good enough with these devices to get a smooth speed graph. That's the reason Doppler speed data is even more important at more than 1 hz.
The speed traces on the sample you posted are typical of the results we saw.
The Doppler speed trace on that sample looks pretty good actually. 
The only issue is that the recording interval seems variable.
Boardsurfr, using a tin can and a watch showing sub-second, walk with the GPS noting your speed, then quickly put the GPS in the can, noting the time on the watch. As a further test, with the GPS covered, turn 90 deg. and walk a bit further.
When you look at the data, you will be able to determine whether the cover-the-GPS-timestamp lines up with the time of the watch, or say 1 second later. And the track and speed's look wierd.
Daffy, we played with a 5hz data-logger (cant remember the chipset), which didn't pass the test above.
Ahh, yes. I remember that now you describe it. I can't remember exactly which it was either.
I went ahead and bought the 10 Hz GPS unit from Ohara. I ignored the fact that the price had gone up by $100 because he added a IMU chip (or maybe I thought I could use those data at some point in the future). Did a few tests - the first one with a bicycle looked great, much more accurate than the GT-31. But when I took the unit out windsurfing, the data looked very bad, with lots of spikes. I learned a few things...
The first thing was that GPS Action Replay will show "Doppler" speeds when you read NMEA files, but that does NOT mean you actually have doppler data. The software simply calls the speed over ground data "Doppler". These data look different from the purely speed based data, but this is just a filter applied. Seems the filter is a bit fancier than just averaging, so it's not immediately obvious. The chip manufacturer (f-tech) has confirmed that the chip does not record doppler. I got a bit fooled by that, and by something I had read that the MKT chip set would support doppler. Finally, I thought that the HDOP and VDOP data the unit can log had something to do with DOPpler - haha.
The unit had a few other problems, too. But with a price of $225 with the IMU chip, and no LED readout, it would not really be an good alternative to the GT-31 even if it had real doppler speeds.
Select to expand quoteboardsurfr said..
... I thought that the HDOP and VDOP data the unit can log had something to do with DOPpler - haha.
HDOP : Horizontal dilution of precision
VDOP : Vertical " " "
ie: in simple terms, the "error" of the GPS engine wrt lan/lon.
I did get a chance to test the FlySight GPS a couple of times now. The unit is based on the u-blox 6 chipset and can log at 5 Hz. Firmware is open source, took me just 30 minutes to make the first custom version (albeit with trivial modifications).
Results look great. The positional accuracy is amazing, the doppler even better. Side-by-side tests with the GT-31 clearly show that the GT-31 has a 2-3 second lag in sudden direction changes (jibes). I plan to blog about, but it's to darn windy here right now :-) But here's a teaser picture from a little bit of playing around in chop:
The FlySight shows the acceleration when going down waves quite nicely, and 5 Hz path shows the direction changes that are completely lost in the GT-31 path. It probably helps that the FlySight can usually track 10-14 satellites, vs. 5-8 for the GT-31. But the filters seem quite a bit better, too.
In jibes, the FlySight also paints a much more accurate picture than the GT-31:
This is what I see in lots of jibes. Seems the GT-31 doppler has a 2-3 second delay, and the positional data keep going forward. Speed and doppler speed disagree a lot. The FlySight data show a path that makes much more sense, and speed is generally very close to doppler speed.
I thought I'd share this before someone buys another GT-31...
