Hi .... and allow for meshes in your synchronization to combat path issues.
Bob On May 13, 2010, at 6:13 AM, Magnus Danielson wrote: > On 05/12/2010 08:47 AM, Eugen Leitl wrote: >> >> Has anyone utilized a network of locally, weakly coupled >> oscillator synchronization (a la >> http://www.projectcomputing.com/resources/sync/index.html ) >> for precise timekeeping purposes? >> > > You have to realize that if you have N clocks interlocked, then their common > frequency will tend to the average or weigthed average (if coupling is not > the same). A single clocks deviation will not have full effect. It will help > to solve phase noise if you monitor the full ensamble of clocks, but this > effect is best at short-term stability. > > It will not solve any form of common environmental effects such as > temperature shifts. Mother earth turns, if faces the sun or the dark space. > AC equipment may first-degree correct for it, but their cycling will replace > it. The environmental effects would dominate iin the time-spans we would be > interested in. > > So in the end, many oscillators isn't as powerful as it may sound at first. > For a network, the solution is to let some nodes have considerable better > oscillators and steer the clock selection accordingly. This is what is done > in SDH networks. See for instance ITU-T G.781. There is a tutorial from ETSI > which designation just slipped my mind. > > The use of mutually synchronised oscillators was investigated in the telecom > world. I don't recall that anything useful came out of that, but searching > for "mutual synchronisation" should get you some hits. I could dig up a few > articles if you care to read about it. I consider it a dead end for the above > reasons. > > Cheers, > Magnus > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
