Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!rpi!julius.cs.uiuc.edu!apple!agate!shelby!unix!hplabs!hpcc05!hpcuhb!hpda!hpcupt1!hprnd!pat
From: pat@hprnd.rose.hp.com (Pat Thaler)
Newsgroups: comp.dcom.lans
Subject: Re: Lightning Protection
Message-ID: <2230132@hprnd.rose.hp.com>
Date: 14 Dec 90 00:29:04 GMT
References: <403@alsys1.aecom.yu.edu>
Organization: HP Roseville Networks Division
Lines: 79

First a disclaimer:  My field of expertise is LANs, particularly IEEE 802.3
and 10BASE-T, not safety.  You should consult a safety engineer about
what is required in the way of surge suppression and fusible links in
your situation.  Also, my opinions are my own and not necessarily 
those of IEEE 802.3.

In comp.dcom.lans, kwe@buit13.bu.edu (Kent England) writes:

(After recommending using fiber instead.)

    	If you insist on twisted pair, you will end up with problems
    due to ground potential differences and the way that Ethernet
    interface manufacturers and PC makers ground their equipment.  In any
    case, it is best not to rely on the end node to deal with potentially
    large ground potential differences, particularly those induced by
    lightning strikes.  It is very expensive to replace lots of PC
    components after every summer thunderstorm.

Walt Haas relates an experience with an RS-232 link which someone
ran between buildings.

I agree with Kent and Walt that it would be best to use fiber optics
between buildings, but the statements about grounding do not apply
to 10BASE-T.  (The question of finding adequate surge suppressors
and the limited distance do apply.)

There is a fundamental difference between 10BASE-T and RS-232.  RS-232
sends single ended signals and a ground between the DTEs at each end.
Differences in ground potentials at the two ends of the link can cause
the link to fail to function.  Larger differences in ground potenetials
can damage equipement or even cause safety hazards.  

10BASE-T sends differential signals with no ground path between the 
equipment.  It does not depend on connected equipment being at a uniform 
ground potential.  Single-ended signalling would be unlikely to be reliable
over twisted pair at this data rate and distance.  The 10BASE-T standard 
requires isolation between the twisted pairs and the equipment which will 
withstand one of the following electric strength tests:

  1500 Vrms at 50 to 60 Hz for one minute (see IEC 435, 5.3.3) or,
  2250 Vdc for 1 minute (see IEC 435, 5.3.3) or,
  2400 V impulse test (10 impulses with peak voltage of 2400 V, 1.2 us
       rise time, 50 us fall time, rep rate of not less than 1 sec; 
       see IEC Publication 60).

These values of isolation are the same as those in ECMA-97 Local Area
Networks Safety Requirements.  (You may want to get a copy of this as
it discusses safety and interbuilding cables.  ECMA is the European
Computer Manufacturers Association.)  This isolation is also on a par
with isolation values generally used in telephone equipement.

Why such high values of isolation?:
Service between buildings was not an objective of 10BASE-T and the distance
of 100 m is short for a between building application.  However, it is
likely for 10BASE-T connections to share patch panels and wire bundles
with phone service.  This implies a possibility of misconnection of 
10BASE-T equipment to wires going outside the building.  Therefore, it
is prudent to require the same kind of isolation necessary for saftey
in case of such a misconnection.

The isolation listed is not enough by itself to provide safety in case
of direct lightning strikes.  It is my understanding that cables coming
into the building should be protected with surge suppression devices
which limit the voltage to 1500 V peak or less and with fusible links
which protect the surge suppression devices from a high power fault
such as a mains fault.  These then limit transient voltages inside
the building to the point where the isolation protects the user and
(probably) the equipment.  (I say probably because a very fast rise
time transient might produce enough differential signal to damage the
receiver.)

At least some surge suppressors have a lot of capacitance.  Connecting
them may cause unacceptable degredation of the signal.  I don't know
if devices are available which will provide the protection needed without
degrading the signal.  This is particularly true when the link is near
maximum length because you can't tolerate much degredation.  If there
are devices which work, I would like to hear about them.

Pat Thaler