Select to expand quotePhromsky said.. Regarding the second part of the answer, how does that work?
Sorry, I'm not with you here, can you be more explicitSelect to expand quote
And which of the two possible speeds that can be arrived at are we obliged to embrace in respect to the GPS Team Challenge?
The team challenge would only like the +/- values to be with acceptable limits, it's not taken into account in the result. The shown result is the most likely outcome. But if you are going for a world record then the terminology is, "at least the lowest number" (somebody may correct me on this).
Well for each of three "2 Second Highs" I counted and added up 10 +/- errors, and on all three occasions they equalled the total +/- error given in the bottom line. So the way I see it is, in all three examples that I looked at, the programme had recorded 10 +/- errors and confirmed the fact in that emboldened bottom line - Strongly implying that the full magnitude of the error observed was founded in all of 10 errors. ..So on what pretext and by what means is the programme able to produce a significantly less error affected and more user friendly score in the result line?
Select to expand quotePhromsky said..
Well for each of three "2 Second Highs" I counted and added up 10 +/- errors, and on all three occasions they equalled the total +/- error given in the bottom line. So the way I see it is, in all three examples that I looked at, the programme had recorded 10 +/- errors and confirmed the fact in that emboldened bottom line - Strongly implying that the full magnitude of the error observed was founded in all of 10 errors. ..So on what pretext and by what means is the programme able to produce a significantly less error affected and more user friendly score in the result line?
I'm certainly no expert, but I think it's because it's assumed the errors cancel out over time.
A lot of the errors are due to atmospheric distortion, a bit like looking through rippled water at an object. As the ripples pass the object will appear to move, so at any one take it could appear to be out of position in one direction, as the ripple moves on, it will appear to be out of position in the other direction. but if you observer it for several cycles you'll get a much better idea of where the object really is. That's why the error goes down with longer times. But as Peter said these results can be overly optimistic. If you look at results from several units mounted next to each other. the differences between them are usually less than the stated result error.
Here's a comparison between my DIY logger and the big motion at 10hz both in my helmet. but speeds for some reason are Km/Hr
You can see how the +/- values decrease with increased time. But for the hour you have to add both +/- figures together to get the compared results within the errors. I don't have the individual +/- numbers handy but they are probably in the region of 0.15 or a bit more
Select to expand quotePhromsky said..
So on what pretext and by what means is the programme able to produce a significantly less error affected and more user friendly score in the result line?
GPS analysis programs like GPSResults and GPSSpeedreader use Gaussian error propagation (typical standard deviation formulas) to calculate the +/- numbers for results. This assumes that the measurement errors are independent (due to "white noise"), which is often true, at least in first approximation. The effect is that the error estimates go down proportionally to the square root of the number of points in a measurement. Thus, 10 second error estimates will be lower than 2 second estimates; 1 hour estimates will always be really low; and 10 Hz data will have lower estimates than 5 Hz data, which have lower estimates than 1 Hz data.
There are numerous complications and exceptions to the general description above. For example, GT-31 data are heavily filtered, so the assumption of independent errors is not valid in the 2-10 second range (but becomes more and more valid the longer the time frame is). This is sometimes taking into consideration by the programs, although I don't think any program gets it perfectly right. Other issues include non-randomness of errors when satellite reception is poor, and differences in how different manufacturers estimate and filter errors.
GPS Speedreader shows the plain average of the error estimates on the bottom line, and this number can be used as an upper limit of the likely error, as long as it is low enough (below ~0.5 knots), and satellite reception was not lost due to a crash or similar. However, none of the error estimates should be taken to draw conclusions about the relative accuracy of devices that use chips based on different designs (Locosys vs. u-blox, like the Motion).
Thanks for your reply Boardsurfr, ..I also came across your article titled "60 Knots - A New Speed record?" ..and so I have appreciated that you are very informed in this matter.
