[ntp:questions] Questions about joining pool.ntp.org

unruh unruh at wormhole.physics.ubc.ca
Thu Sep 1 00:17:02 UTC 2011

On 2011-08-31, Chris Albertson <albertson.chris at gmail.com> wrote:
> positional accuracy is
> On Wed, Aug 31, 2011 at 9:08 AM, unruh <unruh at wormhole.physics.ubc.ca>wrote:
>> On 2011-08-31, Uwe Klein <uwe at klein-habertwedt.de> wrote:
>> > David J Taylor wrote:
>> >> How does this square with those who claim 4ns from their GPS devices?
>> >
>> > Pfft.
>> >
>> > The defining document is rather old I guess. A lot happened in between.
>> > ( I looked into GPS in my diploma thesis ~1987 and not much after that )
>> The GPS sattelites with their onboard clocks, etc are also rather old.
>> And I do not see how  you can get timing accuracies of 2ns when your
>> positional accuracy is 5m.
> positional accuracy is better than 5M.  The Motorola Oncore series of GPS
> timing receivers will perform an automatic  survey.  The reciever runs for
> about 30 minutes or longer and computes it's possition.  You can watch it's
> estimated possition error slowly go down until it is about a foot.

Sorry, I do not believe that. While the random errors do wander around
so that the average is perhaps better than the an individual one, there
are too many systematic potential problems which have a much longer
lifetime than 30 min. 

> Then after the automatic survey the timing error is reduced.  remember that
> for time keeping we assume the antenna remains bolted down and never moves,
>  So after a long period of time the uncertaintly of the antenna location can
> be removed.     Surveyor use GPS and can get cenimeter level accuracy but it
> takes all day or longer to collect enough data.  For nanosecond level timing
> you only need about 1/4 meter and that be be done with  an hour of so of GPS
> data.
> Then along those same lines some people are gets MUCH better timing then 1nS
> by maintaining a very stable local oscillator and phase locking it to the
> GPS' PPS using a very long time constant.   This has the effect
> of averaging out errors in the PPS.     I suspect the GPS itself might be
> long time constant to clean up the PPS.
> You 5M figure, I think might by an instantaneous error circle.   That
> applies to a navigation GPS receiver that can't assume a fixed location.
>  A timing receiver has the ability to average many tens of thousands
> of positional fixes.
> And if (BIG if here) you don't care about the phase but only care that the
> PPS happens at exactly one second intervals you don't NEED to know the
> antenna location but only that it is not moving

The claim is that it is within 2 ns of UTC.
Not 2ns of some unkown time. 

SEe the paper I mentioned where the comparison of one of thse with a
survey grade gps was done. The time offset was over 100ns between the

> One last point.  Almost all timing recivers allow the operator in INPUT a
> surveyed location.  I could hire a survey company to tell my the location of
> the antenn and then remove all uncertainty.   I actually did this, The city
> required me to have a survey before I could pour concrete for a
> house remodel job.  (I had to keep 5 feet clearance from the property line
> and there was question about were the line was)  So I had a survey crew out
> to the house and they worded to the 1/100 of a foot.  So by chance I have

To do that you need a local very nearby survey marker with which you can do
differential GPS, and hope that that marker was actually surveyed to
that accuracy. 
Note that the original lot definition was almost certainly not done with
GPS (unless it is a  very new lot) and thus the lot lines are almost
certainly not defined to that accuracy. You can easily discover that if
you take the lot lines literally, that the lot line runs through the
middle of your living room. 


> means to verify the GPS. (I wouud never have paid for the surveyors if not
> for the city building department insisting on it)  The answer is that my
> cheap $18 Motorola Oncore UT+ can find it's location to about 1/3rd of a
> meter using about 60 minutes of GP data and after that position does not get
> better even after days of averaging.

 Get two of those receivers and see what the time difference between
them is. 


More information about the questions mailing list