On 10/03/2005 3:36 AM, wrote:
I get those clamp on ferrites for a dollar or two at the electronics
fleamarket.

Has anyone ever figured out if it is kosher to put a ferrite on an
ethernet cable. Those cables are real noise sources. I guess they don't
shield them to be cheap or to keep the capacitance low.

Ethernet is actually a number of twisted-pairs. The most common format
is CAT-5E which minimizes most crosstalk (especially near-end variety.)
I think this will almost guarantee some harmonics being radiated.

However, UTP (unshielded twisted pair) cables can radiate *less* than
shielded TP cables. The shield can act as an antenna. There is
industrial stuff wrapped in aluminium that works more like coax or
ladder-line.

How much it radiates is based on frequency, and CAT5E is rated to
350MHz, I think. EMI is actually a problem for these cables. It is
recommended to not run stranded, unshielded UTP cable within 12in of an
EMI source. At any rate, running data through 100baseTx apparently does
not have a real noticeable spike in the noise -- the noise is spread out
quite a bit.

And it's hard to filter EMI on ethernet. A ferrite at each connector
end will not be as useful as on a 60Hz power line. Ethernet is
something like 1v nicely round waveforms. I'm not sure there is even a
line filter available for the frequencies that 100Base-T is expected to
run at.

There are supposed to be filters and/or chokes on specific leads in
switches and NICs, but I'm guessing that they may leave a lot to be
desired. This will only minimize noise coupled to the +DC and GND lines
from the power supply, I suppose.

The rules of thumb I've read a

- keep your runs short as possible
- keep your turns no less than 10x the diameter of the cable
- get or make cables with really good connectors. You don't want any
stretched or untwisted ends -- this is where a lot of the noise is.
- kinks crimps or stretch cables will generate a lot more EMI
- If you have a switch or hub of some sort, but it in a metal enclosure

--
clvrmnky

Heat up and unmunge email to reply.

Good point on shielding the router. Metals boxes and wire mesh for
ventilation are easy to come by.

If I have gear to near to the computer, I can hear a click every two
seconds or so, which I suspect is due to polling on the router, or
communications between the router and the print server.

I have one of those "zap checkers" which indicates RF is being emitted,
though it does not indicate the frequency of the signal. I've seen RF
on the cat5, but haven't bothered to check the boxes.

On 10/03/2005 11:19 PM, wrote:
Good point on shielding the router. Metals boxes and wire mesh for
ventilation are easy to come by.

If I have gear to near to the computer, I can hear a click every two
seconds or so, which I suspect is due to polling on the router, or
communications between the router and the print server.

I have one of those "zap checkers" which indicates RF is being emitted,
though it does not indicate the frequency of the signal. I've seen RF
on the cat5, but haven't bothered to check the boxes.

CAT5 (not CAT5E) has much lesser near-end x-talk protection, and I
recall that it can emit EMI spikes at specific frequencies. This might
be what you are hearing. CAT5E does not have this problem, I understand.

See if the cable and routers are rated CAT5E. It might be worth the
expense to try it out on one data line to see if the noise is minimized.
Heck, it might be worth it to pull most of the cables and see if the
router makes similar noise. The noise may not be network traffic
related. Or power down the router but leave the nodes connected.

Also, another reply to this thread mentions that ferrite material did
choke out common-mode RF. It may help, especially if the router has
less than stellar RF choking on it's ethernet +VDC and GND lines.

If the router is the noisiest part, then my advice about making sure the
connectors are all in good shape still holds. I bet you could snap a
ferrite bead on the power supply/wall wart powering the router, too.

I'm still pretty sure that the twisted-pair lengths themselves are
minimal EMI radiators (or the radiation falls off sharply over
distance), but the ends can still radiate quite a bit, and common-mode
EMI is always a problem. Like most things radio all this is
half-science and half-magic. If it works for you, then it works!

My guess is the better the cable, the better the balance of the twist,
so the lower the RFI. If that were the case, then maybe going cat 6
would be the way.

Also thinking out loud, it seems with DSL you should filter the analog
signals as soon as possible, i.e. where it enters the house rather than
at the phone with those in-line filters. I'm sure the DSL on the crappy
phone twisted pair radiates a bit.

I have a 100Base-T hub and I found that by putting
ferrite cores on EVERY line, power and data, really
quitens things down. The ferrite will stop, or reduce,
common mode emmisions which is the main noise
source. The CAT5 cables are twisted and are
transmission lines with differential RF, and don't
radiate as much, in that mode, as commonly thought.
I am also a ham, and while I don't transmit very often,
I found that my IC28 would cause the router to lock up.
The ferrite split cores stopped that problem as well.
Terry

On 11/03/2005 12:34 PM, wrote:
I have a 100Base-T hub and I found that by putting
ferrite cores on EVERY line, power and data, really
quitens things down. The ferrite will stop, or reduce,
common mode emmisions which is the main noise
source. The CAT5 cables are twisted and are
transmission lines with differential RF, and don't
radiate as much, in that mode, as commonly thought.
I am also a ham, and while I don't transmit very often,
I found that my IC28 would cause the router to lock up.
The ferrite split cores stopped that problem as well.
Terry

Interesting. Of course, the routers and NICs are supposed to use RF
chokes on the +VDC and GND lines to suppress common-mode EMI. If this
leaves some room to be desired (there are no real rules governing how
much RF choking should be done), and you are really using CAT5 (i.e.,
not CAT5E) rated cable and ends, this might explain the interference you
were picking up.

The ferrite beads will not stop the EMI at the frequencies that data is
running down the wire at. As you say, the twisted pairs wouldn't
radiate much.

Common-mode will still be a problem, and ferrite should choke that out.