[time] GPS Time Server Survey

Jan Hoevers ml.janhoevers
Thu Apr 24 22:01:54 UTC 2008


Bj?rn Gabrielsson wrote on 24-4-2008 22:44:
> Hi Jan,
> 
> On Thu, 2008-04-24 at 21:37 +0200, Jan Hoevers wrote:
>> Kipton Moravec wrote on 24-4-2008 18:14:
>>> 1. A GPS module with the ability to expand the serial port to more than
>>> 10 to allow you to use the NTP NEMA driver. The time pulse would be
>>> accurate to about 30 nS RMS. Retail cost about $100.
>> Quite unlikely, since the GPS system itself does not have 30 ns
>> accuracy.  That kind of accuracy is meaningless for pool server
>> operators anyhow.  They typically do not own cesium clocks and the
>> regular pc oscillator is not stable enough to maintain that kind of
>> accuracy.
> 
> I agree with your conclusion that 30ns is meaningless for NTP use.
> 
> However why do you say the GPS system itself does not have 30ns
> accuracy.
> 
> The signal in space (SIS) accuracy quoted for the current constellation
> is between 1 and 2 meters, which is well below 10ns. 

Well, maybe you're right and my information is outdated.
This page (http://tycho.usno.navy.mil/gpstt.html) speaks about "a few
hundred nanoseconds". I recall 340 ns but I forgot where I read it.

Anyhow, a 30 ns accuracy cannot be transferred over an ethernet, let
alone the internet. It cannot be maintained without a cesium time base.

This may be interesting for a well equipped lab, but not for us.

Jan





More information about the pool mailing list