On 14/02/11 00:28, Marco Langbroek wrote: > Op 13-2-2011 02:24, Bill Arnold schreef: > >> This observation was timed with a digital watch that was calibrated with time.gov just before I went out to watch. The time was stated to be accurate to within 0.2 seconds. My watch was in sync with the time.gov page. The watch was calibrated again when I came back in from outside, after about 20 minutes, time remained the same. > > I would strongly advise against using webbased time services for time > calibration. Even if the time server itself is correct, internal processes in > your PC and your ISP's network introduce time lags. > > Time signals obtained over the internet *never* can be accurate. Nothing is ever completely accurate, however NTP manages to hold my machines within a few ms. The protocol is designed to compensate for transmission delays and the systems I use adjust the computer's real time clock oscillator so it keeps very good time for long periods even when the net connection goes away. The software can also use GPS as a reference removing reliance on the net. As you say, one off time requests don't have these advantages, but running a decent ntp client or server referring to several high stratum remote time servers is very reliable unless you require sub millisecond accuracy. Some implementations can also use GPS as a reference removing reliance on the net. The default windows time setting service uses one off requests, often hourly or daily, allowing error to creep in. http://ntp.org/documentation.html _______________________________________________ Seesat-l mailing list http://mailman.satobs.org/mailman/listinfo/seesat-l
This archive was generated by hypermail 2b29 : Sun Feb 13 2011 - 14:30:24 UTC