[time-nuts] Basic Stratum 1 question

[Magnus Danielson]

From: Bill Janssen <billj at ieee.org>
Subject: Re: [time-nuts] Basic Stratum 1 question
Date: Thu, 02 Aug 2007 16:55:21 -0700
Message-ID: <46B26EE9.6090105 at ieee.org>

> ); SAEximRunCond expanded to false
> Errors-To: time-nuts-bounces+magnus=rubidium.dyndns.org at febo.com RETRY
> 
> Richard H McCorkle wrote:
> > ); SAEximRunCond expanded to false
> > Errors-To: time-nuts-bounces+billj=ee.org at febo.com RETRY
> >
> > Jared,
> > Symmetricom has a Stratum definition page at:
> > http://www.ntp-systems.com/think_sync_view_article.asp?ID=42&NewsletterID=1&month=4&year=2005
> >
> > This is what they have to say;
> >
> > Dr. David Mills from the University of Delaware, when he wrote
> > the RFC 1305 that is NTP (Network Time Protocol), developed a
> > hierarchical structure in which Stratum 0 is the USNO (United
> > States Naval Observatory) clock. Stratum 1 is a radio receiver
> > that receives the time from Stratum 0. Stratum 2 is a client
> > that receives the time over a network connection from a Stratum
> > 1 clock. Stratum 3 is a client that gets the time from Stratum
> > 2. This goes on to a theoretical Stratum 15.
> >
> > In the telecommunications world, stratum refers to the holdover
> > performance of an oscillator in the event of loss of
> > synchronization. Stratum 1, Stratum 2, Stratum 3, and Stratum 4
> > are the most typical. A Stratum 1 clock has an accuracy of
> > 1.0 x 10-11, Stratum 2 has an accuracy of 1.6 x 10-8, Stratum 3
> > has an accuracy of 4.6 x 10-6, and Stratum 4 has an accuracy of
> > 3.2 x 10-5.
> >   
> Much NTP stuff clipped
> 
> In the telecommunications world the goal was to prevent buffer overflows 
> (or underflows)
> in equipment that may be connected together even though they might both 
> be Stratum 5.
> In the telecomm. world those buffers are 125 Micro seconds long.
> 
> So the assumptions are made that stratum 5 sync.'s to a 4 and a 4 to a 3 
> and a 3 to a 2
> all the way to a Stratum 1.  In this chain each clock has to follow the 
> frequency wander of the connected lower numbered clock. And it has to 
> have sufficient frequency adjustment range.
> to follow if the lower numbered clock goes into hold-over.

Not quite, the actual network may look like this:

1-3-4-4-2-3-4-3-4-4-3-4-4-3-4-2-3-4-3

P S L L S S L S L L S L L S L S S L S
R E C C S E C E C C E C C E C S E C E
C C     U C   C     C     C   U C   C

Everyone follows (in normal case) the PRC (aka Stratum 1), they are all chained
up. The Line Clock (Stratum 4) in an SONET/SDH switch is always locked to its
SDH Equipment Clock (Stratum 3). For important sites, a Synchronisation Supply
Unit (SSU) clock (Stratum 2) is hooked into the edge SONET/SDH equipment so the
received clock is sent "up" to it before the selected clock is sent back and
the SEC (Stratum 2) of that equipment (or indeed any other edge switch) etc.

There is a very nice diagram of that in the new ANSI spec as I recall it.

Holdover is kept at lowest atainable stratum.

> So there are specifications for each stratum level that hopefully  
> assures stability of the network.

Indeed. Some interesting simulations is the basis for parameter settings.

> BTW the stratum 1 specification came from available Cesium standards 
> used by AT&T
> for many years. They were updated but not by much.
> 
> Been 15 years since I looked at this stuff so it may have changed in the 
> details.

Cheers,
Magnus