POE switches and lightning

Lamar Owen lowen at pari.edu
Fri May 14 12:36:42 CDT 2010


On Thursday 13 May 2010 11:36:35 am Caleb Tennis wrote:
> I was just curious if anyone had seen anything similar to this before?  Our
> incoming electrical power has surge suppression, and the power to the
> switches is all through double conversion UPS, so I'm not quite sure why
> any of them would have been impacted at all.  I'm guessing that the strike
> had some impact on the electrical ground, but I don't know what we can do
> to prevent future strikes from causing the same issues.  Thoughts?

Inductively coupled EMP onto the CAT5.  I've seen ethernet port chips 
vaporized on switches.  I've even seen holes blown in port interface chips, 
and the switch continue working (have a DC powered Catalyst 2900XL switch with 
the center 8 ports in a nonworking state due to EMP from a close strike; the 
2900XL is still running fine, just can't use those center eight ports anymore).  
The building it is installed in is on solar power, and at the time was off-
grid.  A Siteplayer Telnet was blown, and the eight ports were fried (one of 
which was connected to the Siteplayer Telnet that got blown) on the switch, 
but that was the extent of the damage.

I'm from a broadcast engineering background, and have seen lightning's effects 
in many many devices, including vaporized PC traces, etc.  Virtually all 
damage I've seen has been due to either EMP or improperly bonded grounding 
systems.  In particular, if your telecom ground isn't bonded to the electrical 
NEC safety ground, you will get a voltage difference between the grounds, 
depending upon the voltage gradient in the ground.  Whole books have been 
written on this subject; I've got one by Polyphaser about nuclear EMP (same 
concept, larger scale) protection for radio stations.

Imagine the lightning bolt's ionization conduction channel as the primary side 
of many transformers, with every single conductor within many meters being 
potential secondaries.  The closer the secondary, the more coupling.  It's a 
1:1 turns ratio, too, and so a 100% coupled secondary would give an equal 
amperage through the secondary.  Air-core transformers are loosely coupled at 
best, but even a tenth of one percent coupling of a 100kiloampere lightning 
stroke is 100 amps in magnitude.  Loosely coupled current transformers, like 
this, tend to generate large open circuit voltages, too.

The most graphic evidence I've seen of the power of lightning-created EMP was 
made during a strike I saw in June of 1998 at a radio station's studios.  The 
studios were in an old, 1950's vintage school building, built to 1950's civil 
defense standards for EMP resistance (rebar in a Faraday cage arrangement, 
metal roof, lightning rods on the roof).  There is a 100 foot studio-
transmitter link (STL) tower at one end of the building.  The STL tower took a 
direct hit.  The Faraday cage rebar verticals embedded in the walls became 
coupled secondaries, and large currents flowed.

Every single CRT monitor in the entire 300 foot long building was left with a 
rainbow effect on the screen due to the residual magnetism from the EMP.  Even 
monitors that weren't plugged in were rainbowed.  Many PC's died that day, but 
I resurrected several hard drives where I could find identical control boards; 
no hard drive was unreadable due to magnetic issues, but only electrical (no 
bad heads or erased sections on the platters; every one I found a compatible 
replacement control board for was recovered).

Made some good money degaussing CRT's that week.  (used a bulk tape eraser; 
turned on the eraser, brought it close to the CRT, worked it over all 
surfaces, then slowing drew the eraser away from the CRT, and turned it off).

The EMP was strong enough that there were a couple of pieces of spare 
equipment, located in a room less than 30 feet from the tower, that had 
lightning damage even though they weren't plugged in or connected to anything.  
One 250MCM ground wire from the tower was vaporized; there were three, and the 
other two survived, but with noticeable heat-induced discoloration (they were 
replaced, and the glassed-up ground rods were as well).  Engineering estimates 
of the stroke current were that it was somewhat greater than 200kiloamperes.  

One of the STL transmitters was damaged, but on the audio side.  Neither of 
the two STL transmitters sustained any RF output damage thanks to the sacrifice 
of the two daisy-chained Polyphaser arrestors (the arrestors acted as fuses, 
and had to be replaced, but they're a lot cheaper than a 950MHz Marti 
STL-10!).  One of the two four foot Marti STL dishes had a melted feed, but 
the other one, which was lower on the tower (about 85 feet up) was undamaged.  
Fortunately, neither of the two half-inch heliax runs from the dishes were 
damaged.

The 10base-2 LAN took extensive damage, but not every NIC.  The most 
interesting damage was to the RG-58 cable itself, which had holes blown in it 
every 30 feet or so.  Made a good argument to upgrade to 10Base-T at the time.

At my current employer, which is a lightning magnet, we use Altelicon AL-
CAT5HPW lightning arrestors on all cat5 installations that go outside a 
building.  At any building known to have lighting issues, we put one of those 
on every cat5 going to the switch (Altelicon also makes four-port versions).  
Tripplite also makes cat5 PoE compatible arrestors.

Lightning damage is completely predictable, if you have all the information, 
as it's all physics: we just never have complete information, like the 
coupling percentage from the primary ion channel to the various potential 
secondary conductors.  Lightning will take the path of least resistance (which 
may not be the path you think), and it will generate EMP, which will create 
induced currents.  Proper single-point star grounding and bonding of ground 
conductors and electrode fields is a must to reduce damage; multiple electrodes 
or electrode fields must be bonded, or you will get damage.  You may get damage 
anyway; depends entirely on the physics of that particular stroke.

Fun stuff, that's for sure.




More information about the NANOG mailing list