[Robotgroup] A Twisty Question -- re: Cat-5 Cable

[Glenn Currie]

I haven not seen cable specs that specify different twist for adjacent
pairs of CAT-5 cables - (it may well be in the specs) the different number
of twists per meter is electrically significant.

CAT-5 as well as any cable may be considered a transmission line 
and or antenna.  Some of the key circuit considerations for electrical
device include inductance (us ally done with coils of wire), capacitance
(done with conductive plates with an insulator between them) and 
resistance (just a length of wire, carbon or other substance that is less
than a perfect conductor put in line with a conductor). 

These effects are reasonably easy to measure and model in DC circuits.

Once you start to use other than DC you start to have to consider 
inductive reactence ( kind of AC resistance of a coil), and capacitive
reactence ( kind of measure of level of conductance).

So with AC some key characteristics apply, but their effects differ with
the frequency of the signal applied.   So you might think of AC 
resistance as Impedance - that is, frequency dependent.   

For example some loud speakers may be rated as "8 ohm nominal 
impedance" speakers.   This means that as you run audio frequencies from
20 to 20,000 Hz through the speaker - the average AC resistance / impedance
is 8 ohms.  If you measure the impedance at any particular frequency in the
range, you can easily get something different - less than or greater than 8 ohms.

This frequency component of the AC transmission characteristic is used when
building analog crossover networks for multi cone speaker assemblies.
Coils and capacitors are chosen with values to pass low frequencies to the
woofer (large speaker) and higher frequencies to possibly multiple mid range
and high frequency "tweeters:" as the small speakers a called.

So a CAT-5 cable is a transmission line / antenna.   Twisting signal and return
ground wires together does provide significant noise immunity.   If you  use different
amounts of twists in adjacent cables, the resonant frequencies of the pairs will differ and thus be less likely to conflict with each other.

Ethernet uses a 25MHz clock reference in the actual physical signal applied
to the wires.  The pulses are DC but in the big scheme of things, mother
natures says if you start pulsing DC, you have to start playing by AC rules.

The type of cable, the signal and the "magnetics" (line matching circuits) in an Ethernet card are all very carefully 
engineered to to be compatible with the frequency of the signals
applied to the transmission line.

I won't get into all the considerations regarding transmission lines - many
books on that exist and they all get into a lot of math.   The key thing is that
the the components and even the physical shape of the transmission line
(for example twisted or non twisted pairs) becomes important when a signal
is applied.

I think of computer buses, Ethernet cables and radio feed lines and antennas
as variants on a theme.   At the reference clock rate of 25MHz you are way
into the range of radio waves. with Ethernet.  From a basic two way radio
way of looking things you just have an odd antenna and feed line - you 
transmit and receive on an transmission line / antennas that is terminated
on each end, rather then having a transmission line and an antenna in the'
air.  The electromagnetic radiation will make it out of the wire and into the
air to some extent, like it or not.   For a radio you want it to radiate a lot,
for an Ethernet cable you would prefer to minimize the radiation from the
wire and just pick the signal up off then end of the wire with the network interface card.

If are only using one pair in a multi pair cable, the wire with the most
twist will not really make much difference.   The noise cancellation
is mainly relative to the other pairs closely located in the same cable.
To cut out outside noise, you need to add a grounded shield around
the wires.

There a a cost consideration for most cables, so adding extra shields 
is electrically helpful but it costs money.  If you can get away with
just twisting the pairs, that is much cheaper.   For Ethernet cables 
you can get away with it.

Shielded Ethernet cable  is available.

We use Shielded CAT-5 type cable with an outdoor UV light resistant
sheath for our outdoor Wi-Fi type radio work.   Being up on a mast
unshielded CAT-5 can act as an antenna both transmitting and receiving
radio signals.  The UV in sunlight quickly makes untreated plastics fall
apart so you need plastic that is specially treated to last longer in the sun.

The shielding and UV resistance cover s a wast of money in a indoor cable vault.

So, yes different twist on the pairs is electrically significant, and if you are
only using one pair, the one with the greatest twist will not necessarily 
provide any more noise immunity than the others UNLESS you know the
noise is resonant with that specific cable.  In a typical use, it probably makes
no measurable difference which pair you use, as long as the wires in the
pair are intact.

Hope this is helpful, lots more detailed explanations in thick (sometime rather
dull) books.

-Glenn

Joshua Hintze <joshh at imsar.com> wrote: I've heard of that before but I don't have any sources. This is why it is
important however to follow the color specification on the crimp ends. It's
not just so that it has a standard, but it does make a difference.

Josh

-----Original Message-----
From: robotgroup-bounces at puremagic.com
[mailto:robotgroup-bounces at puremagic.com] On Behalf Of Vern Graner
Sent: Thursday, June 19, 2008 1:06 PM
To: The Robot Group Mailing List
Subject: [Robotgroup] A Twisty Question -- re: Cat-5 Cable

During a conversation on network wiring for our office, one engineer and
I were discussing CAT5 specs and how more "twists per foot" would give
better noise rejection.

He mentioned that, of course, CAT5 has a slightly differing twists per
foot density in EACH color pair in order to make sure adjacent pairs do
not conflict. (Huh?)

I had never heard this before and was under the impression that *all*
pairs in the CAT5 cable had the SAME number of twists per foot. A bit of
digging turned up this wikipedia article that tends to support his claim:

http://en.wikipedia.org/wiki/Twisted_pair

  From the above article is this quote:

"Where twist rates are equal, the same conductors of different pairs may
repeatedly lie next to each other, partially undoing the benefits of
differential mode. For this reason it is commonly specified that, at
least for cables containing small numbers of pairs, the twist rates must
differ."

Though the quote states this is "commonly specified", I can find no RFC
or IEEE spec doc that backs up this claim.

So, can anyone point me to a reputable source that can verify or deny
that in a CAT5 cable, the pairs contained (i.e. white orange/orange,
white green/green, white blue/blue, white brown/brown) are wound at a
*differing* twist per foot density?

Also, if it is true that one or more of the four pairs contained in a
typical CAT5 cable are "wound" more tightly (i.e. have more twists per
foot) which ones are they?

This is interesting from the perspective that if there is a pair with
more twists per foot, that cable pair would theoretically have better
noise immunity and if you are going to use fewer than all four pairs,
you might want to favor that pair color in order to get the best signal
quality out of the cable.

Any insight and supporting articles would be appreciated.  :)

Vern

-- 
Vern Graner CNE/CNA/SSE    | "If the network is down, then you're
Senior Systems Engineer    | obviously incompetent so why are we
Texas Information Services | paying you? Of course, if the network
http://www.txis.com        | is up, then we obviously don't need
Austin Office 512 328-8947 | you, so why are we paying you?" CVLG
_______________________________________________
Robotgroup mailing list
Robotgroup at puremagic.com
http://lists.puremagic.com/cgi-bin/mailman/listinfo/robotgroup

_______________________________________________
Robotgroup mailing list
Robotgroup at puremagic.com
http://lists.puremagic.com/cgi-bin/mailman/listinfo/robotgroup