On 12/16/2013 10:25 PM, Jeff Liebermann wrote:
On Mon, 16 Dec 2013 13:42:54 -0500, Jerry Stuckle
wrote:

On 12/16/2013 12:54 PM, Jeff Liebermann wrote:
On Mon, 16 Dec 2013 07:46:16 -0500, Jerry Stuckle
wrote:

Then he has a problem in his ethernet. Properly installed, ethernet
creates virtually no interference.

Yes, that's the theory. If the cable isn't perfectly balanced, it
will become an antenna. Shielded ethernet adds additional
opportunities to radiate. Most of the RFI originates from ethernet
switches and routers.

It's more than just theory. Over the years, we've installed hundreds of
ethernet systems, with little or no interference.

Ummm... you check for RFI problems on hundreds of ethernet
installations? Perhaps I missed that section in the BICSI
certification guides. I don't recall any requirement to test for
excessive radiation, though it probably wouldn't be a bad idea. Good
to see that someone is conscientious about ethernet RFI. (Full
Disclosu I'm not certified but have skimmed the material).


Where it's necessary, yes. And in most commercial installations, it is
necessary. It really doesn't take long, unless it's a huge install.
And no, it isn't part of the BICSI guidelines - but they also aren't the
bottom line in certifications (although they would like to think they are).

The only time I check for radiation is when I install something for
hams. I have a cute little Alinco DJ-X2 HF receiver for sniffing, but
mostly it's just checking with the owners HF radio. Mostly, what I
find is noise originating from the wall warts and power supplies. The
switcher types are awful. I also find some rather noisy LED lighting,
plasma TV's, and computers. Once those are eliminated, the rest of
the noise is conducted by the ethernet wiring. I usually bring an
assortment of big ferrite beads with me to deal with those.


As I said - many commercial installations need it, also.

Shielded Category cable is much better - but you need to be careful. It
needs to be grounded at one end only (to prevent ground loops), and the
shield must carry through to the devices at the far end.

Agreed. Some countries (Switzerland) insist that all CAT5 be
shielded. Too bad that all of the pre-made shielded cables have
shield connections on both ends with no way to break the connection.
If you ground the source end of the cable, you end up with an antenna
and no shielding. Since ethernet has sources at both ends of the
cable, you have to ground both ends to get decent shielding.
See conclusions on Pg 31.
http://dallasemc.org/2010-10EMItroubleshooting.pdf
(Pages 26 thru 31 for ethernet).


No, you do not end up with an antenna if you ground only the source end.
But grounding it at both ends creates a ground loop. This was a
recognized problem even back in the 70's, when I was working for IBM and
we had video terminals on the ends of up to 3000 ft. of coax.

I think you are missing the point here. Terminating the shield at both
ends is not the same as grounding the shield at both ends.

The devices
must be shielded also, preferably in metal cases. Unfortunately, most
consumer-grade devices have plastic cases with little or no shielding
and do not connect to the shield. In cases like this, yes, the shield
can become a radiator. Commercial devices are better at this (but are a
lot more expensive).

Most of the 16, 24, and 48 port switches that I use are in metal
boxes. With all the RFI they generated, a shielded box is required to
pass FCC Part 15. Lately, I'm seeing "green" ethernet switches,
reduce power depending on CAT5 length.
http://www.dlinkgreen.com/energyefficiency.asp
Hopefully, that will reduce radiation by simply lowering transmit
power.


All items which generate rf (which includes basically anything with
digital signals) must pass FCC Part 15. But there are two sections in
Part 15 - Part A (residential) and Part B (commercial), Part A is more
stringent, but you will notice that most devices are Part B certified.
But that also does not mean they will not cause interference; it only
means the interference they cause is within certain limits.

I find the consumer grade plastic boxes are tolerable. Some (i.e.
Linksys) have decent grounding systems. Others have internal shields.
However, I believe that cable shields are detrimental rather than
beneficial. The ability of the CAT5 cable to NOT radiate junk is
dependent on the balance between each wire in a twisted pair. If
anything unbalances the twisted pair, the CMMR is ruined, and the
wires radiate. (Reminder... wires radiate, components do not). Having
3 extra pairs of wires rubbing up against a pair of wires in the
bundle is bad enough for balance. In some CAT6 cables, a cross shaped
insulator runs down the length of the cable to keep the pairs
separated. However, if we add a shield, the capacitance between each
wire of a pair can be different depending on the spacing from the
shield. The change is not much, but sufficient to produce enough
unbalance for the pair to radiate. You can see the problem with a
length of shielded CAT5 and a capacitance meter. Measure the
capacitance from each wire in a pair to the shield. It may vary some
fraction of a PF, but that's enough to unbalance the pair and produce
some radiation. Also, note that it's not necessary for one wire out
of twisted pair to do the radiating. The RFI could easily be radiated
by wires from an adjacent pair if they were sufficiently well coupled.


You can believe all you want. But my experience is much different. For
instance, the presence of a shield actually improves balance in the
wires; the way wires are twisted in a pair averages the distance to the
shield to be the same. A fraction of a pf is not going to significantly
unbalance anything - nothing is perfect, anyway. However, when running
a cable, bringing it near anything metal (even nails in the wall) can
unbalance unshielded category cable, while shielded category cable
minimizes the effects of external metal.

And different pairs have different twists to limit coupling between wires.

One oddity are ethernet hubs. What goes in one port, goes out all the
other ports. With a switch, only data destined for a specific MAC
address on another port is passed. The problem is many older hubs
will send data to a port if a cable is inserted but not terminated.
They're not suppose to do that, but I've found a few old hubs that do
that. If you have an ethernet RFI problem, look into replacing hubs
with switches.

Again, that's the case with consumer-grade goods. Commercial grade are
much better at this.

I beg to differ. Hubs (repeaters) are quite different from switches
(multiport bridges) in operation. The problem is that the hubs will
transmit into a CAT5 cable with nothing at the other end. Common mode
rejection on the pair will prevent it from radiating. However, the
high VSWR caused by the lack of a termination, might cause it to
radiate.


It doesn't make any difference whether you are talking hubs or switches.
Some consumer cheap grade hubs may transmit into unterminated lines.
But better consumer grade and commercial grade hubs will not. They
detect the presence/absence of a device on the other end and operate
accordingly.

"Understanding and Eliminating RF Interference" by Jim Brown K9YC
http://www.audiosystemsgroup.com/RFIHamNCCC.pdf
Start reading at Page 20 for the ethernet radiation section.
Mo
http://www.audiosystemsgroup.com/publish.htm

I think our expertise of doing this as a business for years in both
residential and commercial establishments qualifies us.

With all due respect, that sounds like bluster. You may very well
have had hundreds of successful installations, but I question if they
were all tested for RFI.


I don't really care what you think. And I didn't say they all were
tested for RFI - but a good percentage of them were.

But it sounds like even if we only tested one for RFI, it would be one
more than you have ever done.

This thread may also be of interest:
http://lists.contesting.com/_rfi/2008-11/msg00025.html

Note that 100baseT is quieter than 10baseT. With 100baseT, the data
is first 4B5B encoded at 125Mbits/sec. To reduce crosstalk, it is
then scrambled and MLT-3 encoded. The result is a 31.2MHz carrier and
a mess of sidebands. The necessary scrambling has the side effect of
reducing high power peaks, and evening out the power spectrum over a
wider frequency range, thus reducing the RFI power at any given
frequency. It's much the same idea as the "spread spectrum"
modulation of computer clocks, to spread the power over a wider
frequency range, to meet FCC Part 15 requirements. So, instead of a
carrier or birdie, you'll hear broadband noise, which I guess is more
tolerable.

Again, it depends on the installation and equipment being used. But
10baseT is also old technology. 100baseT is more recent and solves a
lot of problems - both in radiation and susceptibility to interference
from external radiation.

Agreed. Things do depend on the equipment and installation.
Personally, I only do quality installations and use only the best
equipment. For example:
http://www.LearnByDestroying.com/jeffl/pics/drivel/slides/mess01.html
I've seen worse, but the owners didn't want me posting photos.



Low voltage systems have been our business for over 10 years. We
wouldn't have a successful business if we did crap.

And it is our full-time business, not some sideline.

--
==================
Remove the "x" from my email address
Jerry, AI0K

==================