[ntp:questions] Garmin GPS18x - Jitter and adev lab measurements

[David Woolley]

Kasper Pedersen wrote:
> (Initially posted to LinuxPPS. Dave Hart suggested some denizens of 
> questions@ may be interested too.)

Actually, this is news:comp.protocols.time.ntp, questions@ is just a 
gateway.

> 
> These are measurements on a Garmin GPS18x sitting indoors, the PPS fed 
> to a 0.6ns resolution timer/counter. 
> 
> a bit behind, and needs to make an adjustment.
> 
> If you see more than 67ns jitter, it is not the GPS' fault.

The specified time transfer accuracy of GPS is about 50ns; I suspect 
that may allow for some averaging.  If you have a 67ns quantisation 
error, you will also have an additional error due to time transfer 
limitations.

> 
> (For those not used to adev plots: A GPS has poor short term stability,
> but good long term stability, and this produces a downward slope. A

GPS systems need to know the time to a very high resolution to work at 
all.  I think you are referring to either the stability of the PPS 
output, or to the fact that the absolute error is fairly constant, but 
the error relative to the measurement timebase is obviously lower for a 
longer timebase.


> 
> I have measured it to be 60ns+/-100ns slow relative to UTC. 

UTC isn't actually available in real time, although I suspect those 
numbers are still large enough for a real time approximation to be 
defined.  To get the actual UTC time you need to compare various 
national standards after the fact and calculate the difference between 
them and UTC.