Tom brought me down a GT-31/35 today so despite the cold and rain and only about 12 knots of wind we thought we had better give it a go!

This particular GPS unit is a semi-custom made version of the standard GT-31.(Hence GT-31/35).
The main difference is that on top of all the standard features of the GT-31 it also has a MTK 5 hz gps chip built in. To clarify this, yes, it has two gps chips in it. The normal Sirf3 chip runs all the normal functions in the GPS and writes the standard SBP file to the data logger. The MTK chip is separate but uses the same aerial/receiver. It feeds its 5 hz output through the shared central processor for writing to the SD card as NMEA data. Currently, the firmware only provides for the recording of the RMC sentence from the MTK chip due to concerns that trying to write more sentences might overload the processing capacity. This means we have not currently got satellite data or HDOP data from the MTK chip. Tom is confident this will be included soon. The RMC sentence does include location(trackpoints) and speed at 5hz.

I describe this as a semi-custom unit because they are apparently individually hand built but there is some capacity to produce these modified units in very limited numbers for a modest cost premium depending on demand. (Contact Tom if you are interested.)

I will concentrate here on just the 1hz data from the Sirf3 chip for comparison with the GT-11.

The first thing that became clear comparing the GT-11 and GT-31 (1 hz Sirf3 data) worn side by side is that the Doppler speed results were almost identical. The GT-11 speeds were within the range of the two GT-11's worn at the same time.

5 x 10 sec averages:

GT-11 #1: 26.149
GT-11 #2: 26.126
GT-31 (sirf3): 26.128

Only one sample I know but a very good sign.......

When the tracks are overlaid in RealSpeed on the speed graph they are so close they appear to merge into one line unless one zooms in to a very large degree and even then the differences are minimal.
The only difference seen is that the GT-31 speeds are slightly lagging behind the GT-11 speeds on sudden acceleration and deceleration. This is almost certainly a sign of the Sirf3 chip’s improved anti-aliasing filtering which delays the writing of the resulting value by up to 1 second. This should have no effect on the average speeds apart from making them more accurate and less subject to aliasing errors.



Comparing Alphas was even more interesting.

You may have read my earlier comments regarding a strong distrust of the Doppler speeds during tight gybes in Alphas. My opinion after lots of testing is that the Doppler speeds from the GT-11 are fooled into a low spike right in the middle of the turn. The trackpoint speeds seem to reflect much more closely the speed feeling that I get from the actual gybe and display a more realistically smooth speed graph.
This is not the case with the GT-31. The sirf3 chip may have partly solved that issue with the Doppler Alpha speed and the results I got from the GT-31 Doppler Alphas were a bit closer to the GT-11 trackpoints both in average speed and the curve on the speed graph. Unfortunately, the GT-31 trackpoints are totally flawed and clearly incorrect. On most of the Alphas checked there is a high peak just after the slow point of the gybe (just the opposite of the Doppler results from the GT-11). It is immediately obvious and clearly incorrect. Tom speculates that it is probably the result of the filtering being confused by the sudden change of direction (which I think is probably also the most likely cause of the Doppler low spike in the GT-11 Alphas).
Much more testing and many more samples will need to be gathered to get a clearer picture of what is happening in Alphas but my preliminary recommendation for Alpha speed runs is to continue to use Trackpoints from the GT-11 but to use Doppler from the GT-31.
See the screenshot below and pick the odd one out!

To complete the set of Results from yesterdays session:

All speeds are Doppler derived.

5 x 10 sec averages:

GT-11 #1: 26.149
GT-11 #2: 26.126
GT-31 (sirf3): 26.128
GT-31 (MTK 5hz) 26.110

I am a little surprised that the average from the 5hz data is slightly lower just because there are more possibilities to find the very best segment for each 10 seconds but being within .02 knots of 2 others and .04 difference to the other one is pretty insignificant and shows very good agreement.

Alpha speeds from the best run:

GT-11 #1 trackpoints: 18.373
GT-11 #2 trackpoints: 18.341
GT-31 (sirf3) Doppler: 18.01
GT-31 (MTK 5hz) Doppler: 18.007

For comparision:
GT-31 (sirf3) trackpoints*: 19.403
GT-11 #1 Doppler:17.948
GT-11 #2 Doppler:17.968
GT-31 (MTK 5hz) Trackpoints: No result**

* there is an obvious error in this result and almost all of the other Alpha results using Trackpoints in this session from the GT-31 show the same pattern. See image in previous post.

**The trackpoints see saw so much from the MTK data that I have not been able to get a result from either RealSpeed or GPS results, even with filtering turned off.

Note: Trackpoints from the 5hz MTK GPS I have used all show a great deal of see saw or point to point variation. So much that the results seem meaningless.

Example below. Green line underneath is Doppler speed.

I mentioned this some time ago that Higher Hz may not be better for our uses. If a gps has a HPE of 10cm (as the WSSRC requires) and 10HZ ( as the WSSRC requires ) then at 49 knots (25 m/s) instead of having a error of 10cm/2500cm which a 1Hz (ie Navi) gps would record, you get an error of 10cm/250cm every 1/10th of a second. ie 4% instead of 0.4%.
Looking at the graph Andrew posted. There is a 4 knot fluctuation (18-22knots ) over fractions of a second. That is +/- 10% at 20 knots.

Looks like the gt-11 is still the standard.

Trackpoints from Personal GPS even at 1 hz show too much error to be of high value and at 5 and 10 hz they are useless without some sort of filtering for accumulated speed measuring.
On the other hand, Doppler at 5 hz is a very different thing. I don't think the MTK Doppler is filtered and there is still some oscillation on the speed graph but is is relatively small and shows excellent correlation with the 1 hz speed data points (which are filtered). The 5hz Doppler probably shows some real but minute variations in speed due to bouncing and other movements. It is possible that we might be able to see jumps in more detail with 5hz data. The GPS-teamchallenge may be able to add a 'Jump Height' division in the future!

Hi Andrew

Do they look physically different or is it the same shell?? I remember a while ago on GPSS that there was a kind of "wish list" on redesign of the unit (by boogie?) to make it more waterproof to eliminate the need of an aquapac etc. In my opinion it would be great if locosys changed its design so it was more like a big watch than a box. I would imagine the tooling costs wouldnt be cheap though for new moulds etc...

Its great to see guys like you and Tom helping out to make a product like this better for all of us- good stuff!

cheers
lde

Hi Eric, The case of my GT-31 looks identical to the GT-11 apart from the
"GT-31" printing change.

Some other minor changes:
-The screen light is now a light Orange instead of Red and I find this colour soooo much easier to focus on when using it at night (In the car :-)
-The led light now flashes green when the unit is charging and glows continuously green when charged.

Another cool thing on the GT-31 is the ability to flip the display so if you are wearing it upside down you can still read it!

Also, there is a 'maximum time' function designed to be used in conjunction with the 'minimum speed' function. The minimum speed function is not new. It allows you to set a minimum speed below which the GPS will not record data to save memory and useless data clutter. The new 'max.time' function allows you to tell the unit to make a data point even when stationary to a max time limit. In other words, if I set it at 10 seconds, the unit will record a data point at a maximum of every 10 seconds even if it is below the min speed threshold or stationary.
This seems like a good idea for some circumstances like walking back slowly up the beach where you don't want a data point every 1 second but you still get enough data points to see where your have been.

The dual chip 5hz option is now dropped. Locosys have decided not to proceed with this option at the moment and there will be no more 'Engineering Samples' and no firmware support in the near future. As I understand it this was purely a commercial decision based on a perceived lack of demand for this (5hz) configuration.