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

NPG nathan at cmpublishers.com
Tue Aug 30 03:57:06 UTC 2011

On 8/29/2011 9:47 PM, unruh wrote:
> On 2011-08-29, NPG <nathan at cmpublishers.com> wrote:
>> On 8/29/2011 5:03 PM, Chris Albertson wrote:
>>> On Mon, Aug 29, 2011 at 12:43 PM, NPG <nathan at cmpublishers.com> wrote:
>>>> Stupid noob question alert.
>>>> Does GPS provide better accuracy?
>>> Depending on which GPS receiver, GPS can be MANY orders of magnitude better.
>>>   The best GPSes have a one sigma error on the pulse per second that is on
>>> the order of 5 nanoseconds or less.  At this level we need carfully measure
>>> cable lengths and allow for the speed of light delay and "velocity factor"
>>> of the cable.  We are talking about literally 10,000  or even a million
>>> times better, literally.
>> LOL, I only vaguely understand what you are talking about.
>>> That said if the GPS is only being used to drive NTP we don't need
>>> nanosecond level accuracy.    
>> OK, I guess I don't necessarily need to understand the above.
> Not sure what you do not understand, but if nanosecond (1 billionth of a
> second) confuses you then it is not clear what it is you want to do. 

The time stuff I understand, even the radio wave propogation stuff,
though obviously not as well as others here.
I'm not sure what "cable length", "speed of light delay", or "velocity
factor" is all about.

My guess is its about where physics start to limit the theoretical
performance of timekeeping electronics.

> Anyway, GPS can make your computer's time accurate to microsecond (us)
> (1/1000000 of a second) accuracy (It takes time for the computer to
> notice that the gps clock has delivered it once per second signal, and
> to stop whatever it is doing and timestamp that incoming signal so that
> ntp can use it to adjust the clock-- typically of order a few
> microseconds) . WWVB might get you ms (millisecond,
> 1/1000 of a second ) accuracy (because of propagation delays and
> fluctuations, and identifying exactly when the signal came in) 
> . Setting it by hand and ussing something
> like chrony using hand setting could give you 1 second accuracy. 
>>> An NTP server set up well using GPS can keep
>>> time to within about 1uS or 2uS.  A GPS with a 100ns error is good enough.
>>> WWV on the other hand is not so good.  The problem is you don't know
>>> the propagation delay.  The WWV signal has to bounce off the ionosphere at
>>> least once to reach you and you don't know the effective height of
>>> the ionosphere.   Worse, the signal will take multiple paths and what you
>>> get is a mixture,  So you get a "smeared" time signal.  
>> Now that I understand perfectly.
>> :-)
>> Its all about propagation or the lack thereof.
>> Which can be affected by many environmental variables.
>> That probably explains why at least once a year one of my WWVB driven
>> wall clocks loses its mind and starts displaying ridiculous time.
>> :-)
> Probably noise, rather than multipath.
OK, I think I might be using the term propagation incorrectly, although
the end result would be the same.

I thought that a signal rendered unintelligible by noise would be a
propagation failure.
A real propagation failure would be when the signal never got there.

Did I get that right, or just really prove my ignorance?


Nathan Gibbs

Systems Administrator
Christ Media

More information about the questions mailing list