Hi , I have signed Non Disclosure agreements and therefoer I cannot go into what specific manufactures do. But let me say the two or three posts from a couple of people here have expressed what happens.
Rather like the recent thread on power in a heating
element, very few people have the time, knowledge or inclination to understand how GPS works at a first principles level. Hence they make intitutiuve claims or reference articles they either didn't read or understand ( as Dockhead has debunked)
Without descending into the technical issues of pseudo random code correlation , let me state what happens in as non tech as I can make it
(a) ALL gps receiver detect position by detecting the range of the satellite, this is done by inferometry , i.e. phase measurement, By default speed ( and using more satellites , velocity, is determined. instantaneous speed is a function ( note function) of phase detection and relies on the Doppler shift principles.
(b) Instantaneous speed ( I am particular about the term) is not a function of position determination, but as has been pointed out , unless you have a position lock ( in effect a clock correlation ) you havent for a speed. So its more correct to say that both are a function of a lock, but speed is not in itself a function of position determination ( more later)
(c) Because errors in these measurements are distributed randomly or pseudo randomly, simple averaging cannot be sued to remove " noise
" or extraneous results. In simplistic terms , various forms of estimation, are used to in effect determine what errors are outliers. These filters typically use both speed and position to essentially attempt to curve fit the data and remove the outliers.
(d) Instantaneous speed is more accurate then simply using successive positions, simply because the distribution of position errors is such that it would result in a poorer accuracy then using the doppler effect. This is clearly shown in the technical paper referenced previously and in other technical papers.
Remember GPS error positions are statistical probabilities not fixed. Garmin
for example uses the term Estimated Position Error circle, but is often mistakenly assumed by people that all positions lie within the circle , this is not the case.
Now, what a manufacturers determines what you see on the screen
of a GPS, varies all over the place. Garmin
in early models ( upto about 12xl) often showed speed computed from position. However in most cases today instantaneous speed is primarily derived from Doppler calculations
Equally what is stored and output on the NMEA
interface can also be different , Garmin for example never historically stored speed in the track log, until fairly recently.
for the technical minded heres a good paper on the subject http://www.gps-results.com/GPS_Speed.pdf
Note GPS based inferometry is capable of determining heading , including absolute heading while stationary, however its needs two physically seperate antenna
and significant processing power. I have on such device sitting on my desk as we speak.
I really think, unless you have significant technical understanding, debates on such subjects need to be left to people that do. This isn't being elitist , is simply a fact of life.