inter-building ethernet grounding
PAD Powell
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
Dave Martindale
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
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) 852-4769
wunder
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 Natalie <ron>
conduit during lightning storms so just being underground doesn't
solve all your problems.
-Ron
Rod Gilchrist
<>
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
Gerald Scharf
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