As part of that original discussion, there is evidence from the SiRF protocol document in message ID 28, that the unit is capable of 100 millisecond measurements. This could be construed to imply that the sample rate is 10Hz. As to what it can do is still a subject for debate.

sailquik said...
Does anyone really believe that in a tenth of a second the speed accelerated over one whole knot and back again for the next one tenth of a second?

YES

sailquik said...
This is the typical type of noise we see in high hz trackpoints (and Doppler speeds) and must be averaged out over a slightly longer period to get the true picture, which is just what the GT-11 does, hence the lower top speed reported. Dare I say that the GT-11 top speed is more realistic?!

Sorry Andrew, but thats a load of BS. You cant say its noise unless you can verify that with some independently accurate measurement device. This has been done with the Trimble unit. The 10Hz "noise" you see on the consumer GPS is most probably that- scraping down into the noise floor. Sampling at 1Hz reduces the visible noise to 1/10th (since you've slashed the bandwidth by 1/10- assuming no aliasing). Just because your track looks smooth, doesnt mean it is accurate. The high frequency components of motion (clearly indicated as being significant on the trimble data) will simply contribute to aliasing error on the Navi.

I think this comparison is excellent in confirming that the Navi can be capable of producing accurate data at least some of the time.

slowboat said...

sailquik said...
Does anyone really believe that in a tenth of a second the speed accelerated over one whole knot and back again for the next one tenth of a second?

YES

Hmmm. that would be one helluva vibration!!!

slowboat said...

sailquik said...

sailquik said...
This is the typical type of noise we see in high hz trackpoints (and Doppler speeds) and must be averaged out over a slightly longer period to get the true picture, which is just what the GT-11 does, hence the lower top speed reported. Dare I say that the GT-11 top speed is more realistic?!

Sorry Andrew, but thats a load of BS. You cant say its noise unless you can verify that with some independently accurate measurement device. This has been done with the Trimble unit. The 10Hz "noise" you see on the consumer GPS is most probably that- scraping down into the noise floor. Sampling at 1Hz reduces the visible noise to 1/10th (since you've slashed the bandwidth by 1/10- assuming no aliasing). Just because your track looks smooth, doesnt mean it is accurate. The high frequency components of motion (clearly indicated as being significant on the trimble data) will simply contribute to aliasing error on the Navi.

OK, the more realistic comment was a bit tounge in cheek.[}:)]

As I understand it, the verification of the accuracy of the Trimble device is done in a static test over some time to resolve the real position. I don't think Trimble would claim the same level of positional accuracy for each individual point at high speed, but I feel confident that the average of x-n consecutive samples would reach a very high accuracy, that is my point.

Actually, I was thinking of the speed graph sawtooth data from their Trimble GPS that the MI crew published a while back when they blew out their wing. And also the graphs from the survey grade GPS of 20-100hz on the VBox website.

Also, I don't think we are sure that the Sirf3 does not sample at 10hz anyhow. The documentation seems to suggest that at least some sampling/calculations are happening every 100ms. That one needs more investigation.

slowboat said...


I think this comparison is excellent in confirming that the Navi can be capable of producing accurate data at least some of the time.

At least!

I see some places are selling 10Hz gps for $60

www.semsons.com/

GPS Data Logger (SiRF III, Download via Bluetooth and 60,000 trackpoint) for $70

www.semsons.com/glbtbldalo.html

others with 200,000 points ($89).

etc etc.

Locosystems should be able to give us a fully sealed gps with huge memory, bluetooth download, more Hz and induction charging.

wow! - from a simple post this got pretty technical, I dig it!

It would be interesting to install a triaxial accelerometer (probably use one component though) on a board/craft and actually compare/verify the accelerations reported by the GPS.

TimeMachine said...
On the subject of satellites, I believe the doppler result is improved by using satellites low to the horizon, whereas using low satellites can be less accurate for track point location calculations due to greater atmospheric inteference. The SIRF III chips greater sensitivity should help pick these up.

Indeed - however, if the satellites are *very* low, then both gravity and atmosphere will bend the signal somewhat, resulting in decreased accuracy for doppler.

Since this is problem for trackpoint and doppler, I wouldn't be surprised if the GPS's automatically ignored data from satellites when they are low.

TimeMachine said...
Which is another piece of evidence that the data from the SIRF chips is accurate.

Nope- its a piece of evidence that the data from the SIRF chips CAN be accurate. Not IS. To prove something you have to eliminate all possibilities that could potentially bring it unstuck. Its not like "innocent until proven guilty".

TimeMachine said...
the doppler speeds are exported from the GPS to 0.01 m/s in the SIRF binary message but we are reporting it in knots which requires a conversion that can make the result contain many more decimal digits. That is why the results are shown to three decimal places.

That defines the resolution, and its not 3 digits. Converting to knots and expressing with 3 digits actually reduces the resolution slightly. The conversion factor is very close to 2 so the resolution is more like 0.02kts, not 0.001kts.

TimeMachine said...
The SIRF III chips greater sensitivity should help pick these up.

... and all the noise that goes along with greater sensitivity. This is useful for getting some idea of which road a car/pedestrian is on in an urban environment, but does nothing to improve the accuracy of the GPS. The SNR is what determines the accuracy, and increasing the sensitivity does nothing to SNR (it might even make it worse depending on the architecture- depends on what compromise they were willing to take so that their biggest use case is catered for- ie consumer street directories).

sailquik said...

0.01 m/s = 0.019386 Knots

To round it up to 0.02 knots reduces precision.

I was not suggesting that we round it, nor was I suggesting that storing the extra digit was pointless.

slowboat said...
The conversion factor is very close to 2 so the resolution is more like 0.02kts, not 0.001kts.

I was just saying that the underlying data resolution is close to 0.02knots. There will be steps of 0.019386 knots, but you wont easily see them with 3 decimal places. So it might appear that the data has a resolution of 0.001 knots. When you quote speeds in 3 digits, theres a factor of introduced uncertainty in that last 0.02kts, so it becomes irrellevant. ie +- 0.01 knots for each point is all we can get.

Averaging over a number of points (eg 10-sec average) does not improve this, it just reduces the chances of the error being this large. (if you average 10 identical speeds, all reading low, the result is identical to each of the points with the same truncation error). If the data is truncated, at least you can be sure that the measured speed will be lower than the actual speed, but if its rounded you could potentially gain 0.01 of a knot.

I think that as a tool for measuring speeds for training and local competition, the Navi is an excellent device. I have a high level of confidence in the data produced by it (within a fraction of a knot), and its a very big improvement over the old trackpoint-based foretrex.

But it has its limitations for uses beyond this- the main ones I see being:
1) inadequate sampling rate to resolve dynamic events (the Navi is susceptible to aliasing- Manfred's test proved this)
2) The dynamic tracking accuracy has not yet been tested/certified.
3) Susceptibility to external interference has not yet been quantified.

For general sailing around this stuff doesnt matter AT ALL. But for records, I think you have to be absolutely sure that the timing system is thoroughly assessed and quantified before it is qualified for measurement. Thats if the record is to be taken seriously...

I understand there has been a big push to have the Navi accepted for record attempts, and that would be great if its proven to be adequate for the task.

But as yet there has been no independantly certified testing done to assess the dynamic tracking capabilities, nor has there been any testing to prove that we do not introduce aliasing during a record run, when only sampling at 1Hz. The only way to prove this would be with an independant (certified) accellerometer to measure the accelleration profile during the run to estimate the aliasing errors in the GPS log. Its not sufficient to say that "we dont think aliasing will be an issue because we dont think that 0.5G is possible".