[TAG] Wiring a house with ethernet

[Benjamin A. Okopnik]

On Mon, Oct 10, 2005 at 10:28:47PM -0600, Jason Creighton wrote:
> Hey,
> 
> Non-Linux hardware question here regarding wiring a new house with for
> ethernet with Cat5e cable, which I've never done before, so I'm trying
> to be paranoid about it so I don't get anything wrong.
> 
> The concept I have in my mind is to run a bunch of wire to different
> parts of the house from a central location (probably a closet or
> something), crimp standard male RJ45 connectors onto one end and plug it
> into a hub/switch and wire the other end into a female RJ45 jack at the
> other end (bedroom or some other living space) in a nice little outlet
> box, possibly having a phone (RJ11?) jack in the same outlet box, if
> it's a location that needs phone as well.

I agree with Rick that you should be pulling Cat6 if you haven't yet
bought a spool of Cat5; I disagree, however, that you should pull lots
more cable. Instead, pull some 200- or 400-lb. test fishing line along
with your cables, 2x (plus a bit) the total length of the run; every
time you need to add a cable, just pull it from the opposite end of the
run using the fishing line. 

> Questions:
> 
> Am I totally wrong about the basic theory of it?
> 
> Can you run the ethernet cables in the same holes where you've got Romex
> carrying AC, or is there too much interference? 

IIRC, the recommendation is 'at least 18" away from any AC' - but my
memory is a bit shaky. It's definitely worth checking the spec.

> How about with the Cat5
> you've got carrying telephone service? Heck, what about *other* ethernet
> cables?

Those are all fine.

> Which pin-out standard should I use, 568A or 568B?

Either one is good.

> Making ethernet cables is new to me as well. I went out a bought a
> crimper last weekend, and made a little test cable, which *seems* to
> work fine. Are there subtle ways you can get it wrong that will haunt me
> later?
> 
> Is there some easy way to test this stuff without having two laptops at
> the jobsite?

Yes. Again, I agree with Rick that you should use a top-quality,
"automatic" crimper; and again, I disagree that a fox-and-hound (a.k.a.
toner and amp) is a good answer to determining whether the connection's
been properly made. As my lab boss at Hughes once wrote on a big
blackboard after one of the scientists tried to connect two RF circuits
with a wire (and wasted a bunch of time troubleshooting it), "400MHz is
*NOT* DC!" A connection that "rings through" with a fox-and-hound may
look like an open circuit - or a dead short - at high frequency (I
replaced a large number of exactly this kind of cables when I rewired
the Guardian Insurance building in St. Thomas. Say, has anybody noticed
the large number of people who have experienced the joys of cabling
here? I find it rather amusing, actually.)

If you're not buying pre-made or hiring a professional installer, see if
you can rent a cable tester. Conversely, you could rig up two laptops,
flood-ping one from the other, and count lost packets; you should have
some (flooding is like that), but the _ratio_ should remain relatively
constant between individual runs. If you don't do an active test of this
sort, though, you're looking for big trouble.


* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://linuxgazette.net *