It also depends on the hardware and software of the GPS.
GPS uses longitude and latitude to find out where you are, and compares it to last long/lat to find out where you were. The linear distance and the time difference give you the speed.
Do you know this to be true, or are you speculating? For some reason I was under the impression that GPS velocity is based, at least partially, on direct speed measurements through doppler shifts and suchlike. Though maybe some GPS devices don't do this properly; and maybe still others have a GPS chip inside providing them this data, but they don't realise its value and discard it and recompute it the dumb way.
Or, I'm just wrong.
This is a nitpick, of course, it would still suffer from reflections and all that. OBD-II is definitely the preferred solution.
First - Given satellites are orbiting at thousands of miles an hour (14,000km/hour), the addition of your extra few tens of miles an hour in random direction (relative to the satellite orbit) is so insignificant with doppler it will be in the rounding error. Also, doppler (from one source) only works if you are moving in parallel to the satallite's orbit.
A speedometer will give you speed reading even when you are
not moving directly parallel to a satellite's orbit. So, you would still need multiple sats as for sure. Angled motion relative to the satellite's orbit makes doppler shift a trigonometry problem with different shifts from different sats that software must deal with. Could they implement hardware to do it, sure, but it would be nuts to do so from a cost standpoint.
I am not speculating on it having software timing but reasoning - based on experience and reading their manual, I know they are not all alike in how their software/hardware works.
My first has 2 channels, scanned to get the view of 12 satellites.
My second has 4 channels to scan for 12 satellites.
My last TomTom while has one of the most popular chipset at the time, has I don't know how many channels - by then, I didn't care to find out.
All three of them will show signal from all x of 12 visible satellites. With a different number of receivers, they have different hardware. Thus they will have different software and timing.
The TomTom can tell me where it thinks I am when there is only 1 satellites locked (
by looking at TomTom's data stream, I know how many they have locked and the sat data). You need at least two to get actual position (assuming sea-level) and three to get position that is altitude-accurate. So, it stands to reason that the GPS is interpolating. And we all know from experience -- and would guess from experience that a pair of twins writing the same program will likely not have the same bugs. So,
accuracy will highly depends on how their software is constructed.
Some (older) phone-based GPS is
cell-tower assisted triangulation. And some phone GPS was
pure cell-tower triangulation - GPS without satellites at all. This was from phone specs. When my wife was interested in that, I looked at their specs carefully. Useless system if you are out in the country (no signal).
So, it would
depends on how old the OP's old GPS is and how that darn thing works within.
EDIT - Disruptions caused me to lost track of my thought and missed the "why" GPS would
likely have very different correction scheme depending on age and thus position estimation algorithm would likely changed depending on age of GPS.
The thing that made older GPS particularly sensitive to age is changes to the GPS signal quality (by policy, not technology nor science).
When GPS was first opened for civilian use by President Reagan (KAL007 incident had something to do with it), an
artificial error was inserted (~100meter) into the civilian signal. For civilian signal to get accuracy, it has to average over time for a while. The insertion of the error is so civilian signal could not be use for say cruise missiles by regimes not friendly to us.
The inserted error has been reduced and reduced over time. So, older GPS error correction methods (if any) would likely changed rather significantly over time.
If MS Windows is a guide, I bet you enough of them still carries the legacy code that deals with the correction that would have been necessary 5-10 years ago.