[TriLUG] OT: Home Depot and Cat 5

Rob Lockhart rlockhar at trilug.org
Tue Oct 5 20:56:05 EDT 2004


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steve Litt wrote:

| On Tuesday 05 October 2004 09:21 am, Aaron S. Joyner wrote:
|
|> If anyone else has any insights or disagreements, I'd be quite
|> interested to hear.  I am not an authoritative source on cabling
|>  by any means, I've just observed the common industry practices
|> for the better part of a decade and am kind of a stickler when it
|>  comes to proper cabling.
|>
|> Aaron S. Joyner
|
|
| I have an additional question. I have a 150' cat5 cable hand made
| by my vendor. It doesn't work. A friend told me that with long runs
|  there's a special wiring method than with short runs, in order to
|  limit capacitance or inductance or some such.
|
That sounds interesting, until I thought about it.  There's something
called Bob Smith termination that is done on the unused pairs of
ethernet cable, essentially wired as follows:

S W
| |
R R
| |
\/
~ |
~ C
~ |
GND

where S = solid color in pair, W = white with color stripe in same
pair, R = 50 ohms (one each), and C is a capacitor on the order of
1nF.  This is to reduce EMI and keep the long strands from acting like
a common-mode antenna for the two data pairs, and by matching the
characteristic cable impedance, ensure that any EMI is terminated at
each end (instead of being reflected due to impedance mismatch).

Assuming that the isolation transformer has a pretty low capacitance
and thus isolation from ground (or more correctly, from unbalanced or
ground-referenced side of transformer, to balanced side), I wouldn't
think it would matter whether 4 pins or 8 pins are crimped.  However,
if a transformer is not used, or perhaps a crappy one with high
capacitance, then you can see that the circuit above would add a small
capacitance from the unused pairs to ground, and thus add a small
amount of capacitance to one of the differential signals to ground.
This extra capacitance, in theory, could cause extra loading on the
line and act as a low-pass filter which would probably cause the
ethernet mask (eye diagram) to be non-compliant.

The problem with transformers is two-fold; you want good longitudinal
balance (essentially tip-to-ground and ring-to-ground difference) but
low capacitance.  The two work against each other, so no good
transformer has zero capacitance, not without sacrificing longitudinal
balance performnance and thus common-mode rejection ratio.

However, I have used a solid core 8-pair conductor at exactly 328ft
(100m) and had no problems with anything I have tested, for a telecom
project at work.  I would think that not connecting (not terminating)
either end would cause a much larger voltage to be induced in the wire
than terminating.  I believe, due to Lenz' law, this will result in a
larger influence of EMI in the surrounding data-carrying wires, and
thus more detrimental (noisy) to the data.  Yes, I know, they are
twisted pairs, and ideally should only be induced longitudinally
(i.e., equal on each wire of a pair), but there is always some
imbalance in the windings.

In short, I have never seen a problem by NOT crimping all 8
conductors.  Plus, if you ever plan on GigE copper, you'll never do it
with just 2 pairs (all 4 pairs are used, simultaneously, for TX and
RX, at 250Mbps per pair).  The real problem becomes if you have the 4
unused pins crimped on one end but not on another; you can easily see
how that would make an antenna for EMI propagation, from, say, a noisy
power supply or other sources.

Also, FYI, the 100m length restriction is more an issue of timing for
the ethernet collision specs than any EMI / mask problem, from my
measurements anyway.

Hope that helps...

~  -Rob


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFBY0KlcGRHIMeR5bcRAq9/AJ4r1ebpIXiS4xrqN1PuxlqGs2aF7wCgm8Ke
vtRJsg1Tl23HSRjaHpGFg+k=
=whsc
-----END PGP SIGNATURE-----




More information about the TriLUG mailing list