[126411] in North American Network Operators' Group
RE: POE switches and lightning
daemon@ATHENA.MIT.EDU (John Lee)
Fri May 14 16:44:35 2010
From: John Lee <john@internetassociatesllc.com>
To: Caleb Tennis <caleb.tennis@gmail.com>, North American Network Operators
Group <nanog@nanog.org>
Date: Fri, 14 May 2010 15:44:17 -0500
In-Reply-To: <6DECDD46-48BC-46C2-9894-16A0FC4A14BB@gmail.com>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
IMHO,
Long runs of UTP (unshielded twisted pair) make wonderful antenna systems =
for EMI and EMP which is why they are matched to differential drivers and r=
eceivers to reject as much common noise as they are designed to. Older and =
larger Ethernet interfaces have drivers separated from the logic components=
that can handle higher over currents and voltages that are induced on the =
cable. Newer, smaller integral designs cannot usually handle as much power =
as the older designs.=20
UTP systems in industrial environments require higher performance drivers t=
hat handle higher currents and voltages since they are induced by large mot=
ors, HVAC and florescent lighting systems.
EMP (lightening) is a very wide band, high number of frequency bands signal=
with large amounts of power being induced into the systems they are imping=
ing very rapidly. ITGOD (In the Good Old Days) we used to run everything th=
rough conduits which when properly grounded protected both power and signal=
circuits against lightning very well.
There is a very large wifi network in multiple mile long structures connect=
ed by underground tunnels, in the most active thunder storm zone in the cou=
ntry, some of the issues were:
=95 There were a number of power and grounding zones in the buildings
=95 There were local grounds at each of the wiring closets that all equipme=
nt in that zone was tied to. (Have the earth ground checked and if it they =
are corroded or no longer working, have a new earth ground field dug or tie=
to the =93building steel=94 if it has one or more grounding points.)
=95 All inter zone runs were fiber
=95 Both PoE switches and local PoE bricks were used to power the remote ac=
cess points to keep power drops over the utp to within design parameters.
=95 Some zones had switches with both fiber and PoE ports with the PoE port=
s handling local access points and the fiber ports running to smaller remot=
e switches with PoE from there to the edge devices.
=95 If the power runs were too long and their was no local power available,=
custom cables were manufactured to increase power conductor sizes to lesse=
n voltage drops
=95 All outside runs were in conduit and would preferably be fiber.
When I installed my first Ethernet with RG-9, I had to ground the cable at =
the center of the run and tape each end of the cable since it had almost le=
thal voltage at the either end of the Ethernet cable.
My analog circuits professor said forget this digital design stuff, it as a=
n analog signal in a transmission medi(a)um.
Regards,
John (ISDN) Lee
________________________________________
From: Caleb Tennis [caleb.tennis@gmail.com]
Sent: Thursday, May 13, 2010 11:36 AM
To: North American Network Operators Group
Subject: POE switches and lightning
We had a lightning strike nearby yesterday that looks to have come inside o=
ur facility via a feeder circuit that goes outdoors underground to our faci=
lity's gate.
What's interesting is that various POE switches throughout the entire build=
ing seemed to be affected in that some of their ports they just shut down/o=
ff. Rebooting these switches brought everything back to life. It didn't i=
mpact anything non-POE, and even then, only impacted some devices. But it =
was spread across the whole building, across multiple switches.
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 swit=
ches 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 so=
me impact on the electrical ground, but I don't know what we can do to prev=
ent future strikes from causing the same issues. Thoughts?=
