In message <54a8e73d0djcgl@xxxxxxxxxxxxxxx> Jim Lesurf <jcgl@xxxxxxxxxxxxxxx> wrote: snip > Just FWIW and AIUI. > The method changes the clock rate (regulation) to have it run 'slow' if the > clock is ahead, or 'fast' if its behind. If nothing else was then done > later, this would mean it eventually overshoots and runs away. But you also > choose to have system re-regulate at user-specified intervals. The amount > by which the rate is regulated fast or slow is, I think, meant ot be > affected by how much in advance or behind the clock may be. That then means > that as the process is repeated the clock error *should* tend to get > smaller and the regulation 'nudges' smaller. > That should mean that an 'always on' system slowly glides toward having a > small time error and a modest rate correction. > The problem is that this may not work for various reasons, some of which > are obscure. If you're going to test this, also note the times when the > machine is switched on or off. At the moment I'm running it (ARMiniX) 24/7 and running *nettime_status (from time to time) in a task window which I can save. Once I have sufficient data I will upload the results to my website. Initially it was 20 seconds slow. It's now within 1 second (when I compare it with my PC which runs ntp). It will be interesting to see how much "drift" occurs after a shutdown period. Barry -- www.g4iat.co.uk Using an ARMiniX with RISC OS 5.21 SP7b (21-Apr-14) http://www.g4iat.co.uk/wx/noaa/ http://www.wunderground.com/weatherstation/WXDailyHistory.asp?ID=IENGLAND203 http://app.weathercloud.net/map/ --- To alter your preferences or leave the group, visit //www.freelists.org/list/armini-support List-related queries to info@xxxxxxxxxxxx