[54682] in North American Network Operators' Group
Re: fast ethernet limits
daemon@ATHENA.MIT.EDU (Kevin Oberman)
Mon Jan 13 11:34:16 2003
To: Måns Nilsson <mansaxel@sunet.se>
Cc: nanog@merit.edu
In-reply-to: Your message of "Sun, 12 Jan 2003 19:59:08 +0100."
<108530000.1042397948@localhost>
Date: Mon, 13 Jan 2003 08:33:26 -0800
From: "Kevin Oberman" <oberman@es.net>
Errors-To: owner-nanog-outgoing@merit.edu
> Date: Sun, 12 Jan 2003 19:59:08 +0100
> From: =?ISO-8859-1?Q?M=E5ns_Nilsson?= <mansaxel@sunet.se>
> Sender: owner-nanog@merit.edu
>
> Some hours reading the back issues of the journal found at
> http://www.compliance-club.com will hopefully inform you why star grounding
> is a thing of the past. Ground both ends. If you are afraid of ground
> loops, place a heavy (as in 10-16mm2 or AWG way below 10) ground conductor
> alongside the signal cable, and ground it firmly in both ends. That will
> take the current away from the shields.
>
> Ungrounded shileds are inefficient for EMI and RF shielding, while at times
> efficient AC hum blockers.
>
> And, IANAEE, but I've played with big sound systems that exhibit all these
> problems.
I find this simply frightening! Have you any idea how much potential
of ground can vary in a large building? It's easy to have AMPS of
current flowing through the shield of a cable and enough voltage
offset to be dangerous to people. (802.3 allows enough breakdown
potential that the equipment is unlikely to have a problem, though.)
Grounded at one end is better than floating for shielding, so this
almost reasonable (as long as the wiring is all installed to spec) but
grounding at both ends in the wrong environment can lead to serious
problems.
Since the 802.3 sections on 10Base-T does not deal with shielded
wire at all, there is nothing there on the subject. But other sections
on 10Base5 and 10Base2 are explicit that 10Base5 MUST be grounded at
exactly one point and 10Base2 recommends that one point be grounded
when the cable extends out of a room. More than one ground is
explicitly prohibited.
The archives of comp.dcom.lans.ethernet are full of people who have
high error rates because of multiple grounds.
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman@es.net Phone: +1 510 486-8634
