> I believe you may have missed the "smoking gun" or maybe I
> don't quite
> understand how you did the tests. It appears that the
> Netgear hub, not the
> router, is the primary source of your problem. You demated
> the cable from
> the router end and the noise dropped significantly
> indicating that there is
> likely (common mode) noise from within the hub being
> coupled onto that
> cable. I'd recommend trying the test again of applying a
> ferrite to that
> cable at the hub that will give you several hundred or a
> thousand ohms at
> the frequency of interest. In my experience, the common
> mode noise is often
> coupled out of both the data cables and the DC power
> cable. Try ferrite
> loading all cables at once on just a single box.
Think of the offending device as a transmitter and the
attached cables as parts of an antenna. Also consider the
frequency. Looking at things this way gives you a feel for
what is happening.
For example, direct radiation out of a box on HF is very
low unless power levels are very high. The reason is the
antenna area of the box is too small to be an effective
radiator. Also a metal foil "shield" laid on the box won't
do anything, it is too small in terms of wavelength to be a
reflector and it isn't a closed box or waveguide.
Now consider what you actually do with ferrites. All you do
is add a little common mode impedance, the exact amount
depending on the material. A 73 mix material roughly adds
about 50-100 ohms per linear inch of ferrite surrounding the
cable at 2MHz. Make two turns through that same core and you
have 200-400 ohms at 2MHz.
Depending on the impedance of the source inside the box and
the common mode (longwire antenna mode) impedance of the
cable or wire leaving the box, that added impedance might
make a huge change or it might make no change at all. The
added impedance can even make radiation worse if it happens
to match impedances between the source and "antenna"!
What we tend to do is ignore how the system actually works
and imagine we are doing always something significant and
constructive just by throwing an unknown impedance into an
unknown complex impedance of source and loads. "Well gosh,
this 600 ohm impedance I added works on 400MHz in that
system, why doesn't it work on 4 MHz in that system?" The
answer is not only because the SYSTEM impedances are totally
different, the same parts behave totally different with
Imagine a large antenna made from multiple wires running
different directions that is broken at one point and excited
by a low impedance source. Say the exciting source (unwanted
transmitter) impedance is very low, and the antenna (router
cables) have a feedpoint impedance of 20 ohms. The problem
is at 4 MHz.
We throw a chunk of foil over the top of the five inch wide
transmitter. While that foil might be a perfect reflector at
900MHz and the box a significant sized antenna at that
frequency beaming the unwanted radiation down into the
floor, it is absolutely insignificant at 4MHz.
Now we add a bead that adds 100 ohms of resistance in series
with the 20 ohm antenna feedpoint and we see a significant
decrease in unwanted antenna currents.
On the other hand if the lines leaving the unintentional
transmitter look like a 500 ohm impedance longwire or
dipole, the very same beads will do next to nothing. If the
unintentional transmitter likes to see a 550 ohm or higher
impedance load, the very same beads can increase radiation!
It's the system that matters, and every system is different.
Unless you measure things, it's always cut and try or just
pure dumb luck.
Once I learned what device was causing the problem, I'd put
a very high common mode impedance at the offending frequency
on every line leaving that point. Not just some piddly
little 100 ohm snap on bead that works through dumb luck in
some cases. If that didn't do it, I'd look for alternatives
like common mode choking the lines at several points.
This is why shields and metalic conduit are so nice. You can
ground them or bond them or isolate them at junctions to
greatly alter common mode impedance without upsetting
desired signals inside the enclosed cables.
> Remember, this is a "war of attrition." For example,
> assume that you have
> two cables that are contributing equally to the total
> power being emitted by
> the noise source, if you eliminate one of them, your noise
> level will only
> decrease 3 dB (very roughly 1/2 S-unit).
Radiation from an additional path or paths might increase
field strength or very well might decrease it. The level can
change by almost any amount in almost any direction or
distance. It's a very complex interaction of peaks and nulls
as fields combine.
The unfortunate thing about this is much of the solution is
found threw the Edisonian method. Cut and try that is sped
up with a little intuitive "feel" for the problem so you
don't waste time with obvious non-solutions.
RFI mailing list