[TowerTalk] Lightning and wired networks was: Re: Lightning protection

[Jim Lux]

On 7/9/13 6:40 AM, Pete Smith N4ZR wrote:
> Sure, but the network cables were connected directly to the computer
> motherboards.  I have also lost a couple of RS-232 ports on my K3, which
> now has a DB-9 surge protector on it - I'm inclined to think it's a
> design problem.



Interestingly, ethernet should be easy to protect: it's galvanically 
isolated by a transformer, it's twisted pairs, etc.

It would be interesting to see what the failure path was.  Was it a 
common mode current induced in all 8 wires that raised the voltage above 
the few thousand volts the isolation transformer handles?
(standard spec is 1500V rms, higher is available, as are other 
protection devices)

http://www.digikey.com/Web%20Export/Supplier%20Content/Bourns_118/PDF/Bourns_ProtectingEthernetSolutions.pdf?redirected=1 
  has some discussion of this

Littlefuse also has a nice ap note..


I noted in my googling that some Ethernet devices do not use a 
transformer, and have no isolation (raspberry pi, apparently)


or was the capacitive coupling across the transformers at either end 
(presumably there's some sort of switch or hub at the other end which is 
also transformer coupled) enough to carry enough energy to do damage.

If there's Power over Ethernet involved that also gets in the mix.

I would think, though, that there is fairly robust Ethernet available, 
since a LOT of ethernet is installed in lightning areas, including 
outdoor installations of shorthaul microwave links, WiFi access points, 
etc.   The inexpensive $20 switch for consumer use might not be as 
robust, but at that price, maybe it could be considered a consumable fuse?

---


The serial ports are another matter. RS232 was never intended for long 
runs, of course. And it's not galvanically isolated, although in a good 
implementation, it can take a pretty healthy voltage AND source/sink 
30mA on any pin.  But a lot of serial interfaces bring the TTL signals 
out directly with no buffering or they don't use clamp diodes (because 
the capacitance of the diodes limits the maximum speed) or.. tons of 
reasons.

The vast majority of RS232 installations also do not use a shielded 
cable with the shield tied to chassis, which is separate from the pin 
7(25 pin connector) signal common. So induced voltages between boxes are 
a big problem with RS232. (not even getting into the "circuit common vs 
chassis ground" issue)

(Those telegraphers and teletypes with 60mA current loop signalling 
weren't fools.  Nothing like miles of wire strung across the countryside 
to be a transient target and promote "robust" equipment design.)