[time-nuts] Noise and non-linear behaviour of ferrite transformers

[John Miles]

> > All that said, the real hazard with transformers is that people tend to
use
> > them to drive unbalanced coax cables with balanced signals.  This turns
the
> > coax shield into an antenna, at which point you may end up with with
more
> > noise and spurs than you had before.
> 
> Could you explain this a little bit more? Because this would be exactly
> what i would like to do.

I often find that when I use coaxial baluns to cut down on ground loop
noise, I end up with more noise and interference than I started with.  Not
always, but often enough that I'm leery of them.  

Due to skin effect, most signal propagation in a coaxial cable takes place
between the outer surface of the center conductor and the inner surface of
the braid.  Ideally, the outer surface of the braid just underneath the
jacket will act like an equipotential shield to keep external EMI away from
the signal path inside the cable.  

But that's really only true when the cable connects two devices in
well-shielded enclosures that are themselves at a similar ground potential.
When you "lift the ground" with a coaxial balun such as an FTB-1-1+, you can
no longer pretend that the coax braid is at ground potential along its
length.  From an RF perspective the braid is floating at one end, which
makes it an antenna.

Put another way, a balun will reject common-mode signals in favor of
differential signals.  That's fine if you're using it with a twisted pair or
other balanced line (you're probably aware that this is how UTP Ethernet
cables work).  RF interference in such a line is picked up equally by both
conductors and rejected by the balun.  But a length of coax cable is as far
from a balanced line as you can get.  One conductor is well-shielded, the
other has its outer surface flapping in the breeze.  The balun can't tell
the difference between desired signals on the inside surface of the braid
and undesired signals on its outside surface.  They both look like
differential-mode signals, relative to the inner conductor.

The same thing happens with instruments that allow you to lift the ground at
their input jacks.  Apart from the unwanted-antenna effect, this is almost
always a bad idea because it's very hard to properly ground the jack's outer
shell to the chassis.  Few things in EMC are more important than ground
integrity at the point of entry to an enclosure. 

When fighting ground loops, a good first step is to minimize the loop area
if you can.  Try plugging your DUT, reference, instrumentation, and computer
into a single power strip.  That will take care of most of your power-line
interference problems.  Baluns can help too, but don't be surprised if they
don't.

-- john, KE5FX
Miles Design LLC