[time-nuts] Basic Stratum 1 question
magnus at rubidium.dyndns.org
Thu Aug 2 14:40:04 EDT 2007
From: John Ackermann N8UR <jra at febo.com>
Subject: Re: [time-nuts] Basic Stratum 1 question
Date: Thu, 02 Aug 2007 11:56:51 -0400
Message-ID: <46B1FEC3.5020905 at febo.com>
> ); SAEximRunCond expanded to false
> Errors-To: time-nuts-bounces+magnus=rubidium.dyndns.org at febo.com RETRY
> Jared Morrisen wrote:
> > ); SAEximRunCond expanded to false
> > Errors-To: time-nuts-bounces+jra=fluffles.febo.com at febo.com RETRY
> > Hi,
> > I am having a debate with our CIO. He wrote in a memo about timing:
> > *Local hardware is to be considered Stratum 1, since it get time from its
> > own CMOS.*
> > I told him that absurd and that it can't be considered stratum 1.
> Hi Jared --
> I don't want to get in the middle of that argument!
> However, note that there are two different definitions of "Stratum 1"
> floating around.
... at least, and within the synchronisation community. The stratum as such
only significance a hierarchial order.
> In the NTP sense, it is nearness to a reference clock that
> (theoretically) provides time traceable to a national institute. So an
> NTP "stratum 1" server is one that is directly connected to a reference
> clock. (Check the NTP website; I'm sure you'll be able to find a more
> formal version of that definition somewhere there.)
The NTP Stratum meaning is really a hop-count measure, a metric.
> But in the telecom industry, "Stratum 1" signifies a certain level of
> timing performance, and I suppose that some computer system somewhere in
> the world might be able to meet that standard for some period of time.
> Someone else will be able to tell you just what the definition of
> Stratum 1 is.
The SONET Stratum numbers indicate the level of clocks being interconnected:
Stratum 1 - Primary Reference Clock (i.e. Cesium, Hydrogen or GPS) +/- 1E-11 in frequency relative UTC
Stratum 2 - Station clock (i.e. high quality OCXO or Rb-cells)
Stratum 3 - Equipment clock (i.e. good OCXO or very good TCXO) +/- 4.6E-6 in frequency
Stratum 4 - Line clocks +/- 20E-6 in frequency
You always (in the ideal) try to hold-over from the lowest stratum, i.e. the
highest stability. A lower Stratum clock does not listen to a higher stratum
clocks holdover, but it will certainly listen to it when it is locked directly
or indirectly to a clock of same or lower stratum number.
These Stratum concepts are quite different and it is unfortunate that they
coexist and help confusing the issues. An NTP Statum 1 clock can be either a
very high quality one or one of very low quality. The NTP Stratum number says
nothing as such about stability and correctness in tracking relative UTC.
> So the real question is which of those definitions is your CIO using?
> Based on that, you should be able to answer that either (a) local CMOS
> isn't a reference clock traceable to a national institute, or (b) that
> the CMOS clock isn't certified to meet telecom Stratus 1 standards.
> Of course, if his point is that one local free-running clock is the
> "master" and that he cares only about synchronization, not accuracy,
> across the network, then he might be making sense. But he's not using
> the terminology correctly.
Indeed. For some purposes all you need is to have the same time scale,
regardless if that matches the surrounding world or not.
More information about the time-nuts