johnsson@decwrl.UUCP (Richard Johnsson) (08/24/84)
The Xerox Palo Alto complex of buildings had a lightning induced Ethernet
failure several years (which probably led to the emphasis on grounding in
the Blue Book).
Basically the cable became grounded in two buildings about 500 feet apart
by cable taps accidentally touching metal cable trays. A nearby lightning
strike induced a large, temporary potential difference between the buildings
and fried several tranceivers. Unfortunately about six of the tranceivers
died shorted rather than open and we had a lot of fun learning how to
use a TDR to find them.
Richard Johnsson
DEC Western Software Lab, Los Altos, CA (415) 949-1494
{decvax,ucbvax,ihnp4,allegra}!decwrl!johnsson
johnsson@DECWRL.ARPA
RHEA::JOHNSSONpadpowell@wateng.UUCP (PAD Powell) (08/25/84)
I think that you can stop worrying about lightning strikes. Usually this occurs when the cables are EXPOSED, as in "hung up on a tree". I would suggest trenching and burying your cables. If you are concerned about ground potentials, why not simply run coax, and put a couple of isolators at critical points, such as the nearest building entrance? I would strongly suggest tying the ground (shield) of the cable at that point to a REAL EARTH (read- a big metal pole in the ground). By the way, somebody made an "isolating repeater" at a very resonable price, but I cannot find it in my files. I was really concerned about this problem in the past, but other ones seem to be much more critical, such as "tromping on the cable", installers who tie kinks in it, etc. Patrick Powell
dmmartindale@watcgl.UUCP (Dave Martindale) (08/25/84)
Direct lightning strikes will, of course, wreak havoc with an ethernet cable. But I believe previous respondents' concern with lightning strikes was that a lightning strike on one building could momentarily raise that building's AC ground potential considerably above that of nearby buildings, thus causing ground currents between buildings via the Ethernet cable if the cable was grounded in both buildings. The solution, of course, is to ground the Ethernet cable in one place. But even then, the ground potential difference could cause the normal isolation built into the transceiver to break down. (Anybody know what the minimum isolation is?) And, of course, anyone working on an Ethernet cable in one building when the cable is grounded in another building and there is a thunderstorm in progress might get a nasty shock.
wunder@wdl1.UUCP (wunder ) (08/26/84)
We are facing a similar problem here (Ford Aerospace), and don't
have a good soloution either. I have thought of two things, though.
Thing 1: IEEE 802.3 requires transceiver cables to be isolated from
the coax. This means that each 500m Ethernet segment should (must?)
be individually grounded. This fixes the ground loop problems.
>From the standard (Draft, Rev. D, December 1982):
"7.4.2 MAU Electrical Characteristics
7.4.2.1 Electrical Isolation
The MAU must provide isolation between the AU Interface cable
and the coaxial trunk cable. The isolation impedance shall be
greater than 250kohms, measured between any conductor (including
shield) of the AU Interface cable and either the center conductor
or shield of the coaxial cable, at 60 Hz. The breakdown voltage
of the isolation means provided shall be at least 250 VAC, rms."
An MAU is a transceiver (Media Access Unit) and an AU Interface cable
is a transceiver cable.
Thing 2: You should be able to provide lightning protection with
lightning arrestors where the cable enters the building. Radio receivers
are at least as fragile as Ethernet transceivers, and they have been
protected for years. I don't have any ready references for lightning
arrestors, but I believe that GE makes them, and Andrews (makers of
antennas and transmitting coax) should be able to help. With arrestors,
you should be able to run regular coax between your half-repeaters
instead of fiber optics (talking about the runs between buildings, here).
It should be possible to get an arrestor that would go right in the
Ethernet coax -- all you have to do is keep the impedance within specs.
Good luck,
walter underwood
UUCP: fortune!wdl1!wunder
ARPA: wunder@FORD-WDL1
Phone: (415) 852-4769wunder@wdl1.UUCP (wunder ) (08/26/84)
Oops, one more thing. General Electric (or Westinghouse, or ...) should have lots of data on how likely a lightning strike is in your area. They use that info to sell lighting arrestors. w underwood
ron@brl-tgr.ARPA (Ron Natalie <ron>) (08/28/84)
We've had serious problems with cables that are underground and in conduit during lightning storms so just being underground doesn't solve all your problems. -Ron
rod@aesat.UUCP (Rod Gilchrist) (08/29/84)
<> About that 1,000,000 volt transient .... There is a body of knowledge out there related to lighting strikes. In particular the problem has cropped up in both telephone and power transmision industries (ever notice that your telephone doesn't seem to have much trouble with lightning, even though the wires are unsheilded and hung up in the air?). I believe the initial telephone papers were published in the Bell Systems Technical Journal in the fifties or early sixties describing work done to measure lightning induced transients (sorry I can't be more specific, its been a while since I was involved in this). What was discovered is that although there is the occasional induced voltage that was unmeasurable because it destroyed the measuring equipment, in the very large majority of the cases the transients were less than 2000 volts. The numbers I recall are that 95% of all strike induced transients on (buried) coax cable were measured to be less than 700 volts, with a duration of in the order of 100 microseconds (they were more like 1500 volts with a shorter duration on twisted pair). One result of the work done was an IEEE standard for power system measuring equipment that requires that signal wires be capable of with standing about 50 one microsecond, 2000 volt spikes one millisecond apart. This is intended to insure that it is very unlikely that the equipment will be damaged by lighting or other transients. I'm not familiar enough with ethernet transcievers to relate the above to them, but I thought I would pass this much along. --- Rod Gilchrist @ AES Data Inc 1900 Minnesota Court Mississauga, Ontario, L5N 3C9 416-821-9190
scharf@cornell.UUCP (Gerald Scharf) (08/30/84)
<>
There must be people out there who have ethernets, or for that matter
any coax based net that spans several buildings. As one of the technical
people within Cornell designing the net, I am asking for the input of other
net designers.
In particular, an environmental engineer from DEC has proposed one
catastrophic failure syndrome. If there is a electrical storm with a strike
near one end of the coax, there will be a temporary induced ground potential
in the order of millions of volts relative to the cable ground. If there is
sufficient voltage, the insulating jacket of the cable will break down and
arc to ground. This fries the cable instantly, probably takes several
transceivers with it, and could even fry the repeater logic boards. Wow.
We decided that trying to tie all the building grounds together with
00 or 000 wire (read this as size of a thumb copper wire 1300 feet long,
read get a major NSF grant to pay for ground connection) was unacceptable. So,
the other proposed alternative is to put the coax backbone in one building,
just long enough to connect 10 transceivers, then use remote (fiber optic)
repeaters to connect to each building coax segment. DEC says that they can and
have connected several remote repeaters in one net, despite Blue Book
restrictions. The unfortunate side effect (for us, not the vendors) is that
it doubles the total cost of the network before building segments to almost
$60000. I am also having trouble locating a supplier who can supply almost
2000m of 2 conductor 100/140u fiber able to withstand freezing, in less than
6 weeks.
Has anyone ever had a net fried by lightning? How can I determine the
number of times in a ten year period that this is possible? How does one factor
net availability into cost justifications for insurance/safety of the net?
Do other people have uneasy feelings about using a bunch of remote repeaters?
Jerry Scharf
no longer ignorant, I seek knowledge not enlightenment