Actually a gps provides only a position signal not a speed signal and the devices software has to deduce the speed out of it, which requires to steps:
>translating position signals in WGS84 to distances. This is more tricky since the earth is not sphere (as asumed in many devices) nor an ellipsoid. But the error here is small.
>differentiating the distance over time. Sounds simple, but because the position signal is noisy, the error is amplified by the differentiating.
Gahh...
GPS velocity measurements are based on dopler shift NOT differentiation of the position data. As such they are a LOT more accurate at measuring velocity than they are at measuring position.
My first two google hits actually say velocity estimates are usually within 1cm/sec (0.02 knots!).
simple explanation (from an aircraft forum)and
A technical peer reviewed scientific paper You are right though in that there are other huge sources of error, like sailed track length (in 3D!) over which you are averaging.
Personally I think it's amazing that GPS works at all... they have to take into account the relativistic effects of gravity on time dilation to make it work!
