[ntp:questions] Re: May ntp be off by a second in case of GPSsource?

Richard B. Gilbert rgilbert88 at comcast.net
Sat Dec 31 01:38:28 UTC 2005

Eric wrote:

>On Fri, 30 Dec 2005 19:47:28 GMT, KomkaP at prolan.hu (Komka Péter) wrote
>for the entire planet to see:
>>>NTP doesn't require a local reference clock at all.
>>I meant not "cannot do without a GPS" but "(in case of GPS) cannot do with 
>>*any* type of GPS". (It does matter the behaviour of the GPS. To which PPS the 
>>stamp belongs (if either).)
>>(NMEA itself does not specify that?..)
>Well, I'm not sure if I understand your question, but there are many
>types of GPS receivers, and NTP can interface with more than one type,
>but probably not all types.
>Some GPS receivers seem to be adapted for timing functions, but there
>are many that are not good for timing, only for location.  NTP doesn't
>know or care where a reference clock gets its time from.  GPS, ACTS,
>CDMA, rubidium, etc.  

GPS receivers are generally optimized for either timing or position.   
It is difficult to do both well at the same time and very few 
applications require both at the same time.  So, when selecting a GPS 
receiver for a timing application, you must be sure to get a version 
optimized for timing.   Any GPS receiver knows very well what time it 
is; it solves four simultaneous equations in four unknowns; latitude, 
longitude, altitude and time.   The issue is the presentation of the 
data to your computer.

The timing receivers have a pulse per second (PPS) output.  They 
generally do a "site survey" when first installed, to figure out exactly 
where they are.  Having located themselves, they lock in the position 
and subsequently report only the time.

More information about the questions mailing list