[time-nuts] Question on GPS receiver selection for 1pps

[WarrenS]

Paul

All depends on how "Nuts" you want to be and how inaccurate the thing is that you are locking.
It comes down to how long you want to average over and how good do you want your frequency to be.

The more the GPS noise the longer you have to average to reduce the effect of the noise.
This is generally is done by slowing down the control loop time constant.
BUT the slower the control loop, the more the oscillator is going to drift over that time period before it is pulled back in.

Over simplified  example.  If you only care about long term freq or phase comparisons say over many hrs to days, then the noise does not mater, (if it is handled correctly)
If you want low noise of 1e-12 over say 1 to 1000 sec tau, then  you're going to need about 1 ns.
If a frequency noise of 1e-9 is acceptable up to the 1000 sec range then 1us noise could be made to work.

ws
*******************

[time-nuts] Question on GPS receiver selection for 1pps
paul swed paulswedb at gmail.com 
Thu Nov 18 16:05:55 UTC 2010 


OK interesting but the question was what is the need for a reasonable 1 pps
for gpsdo locking. The first comment on this is anything below 30 ns.