Bill Sloman said:
A balun is actually a transmission line transformer.
Not a necessary construction method; a balun is just a transformer with
tapping such that it inverts one side.
The twisted pair that constitutes the bifilar winding is a
transmission line, with a particular characteristic impedance which
depends on the diameter of the wire involved and the thickness and
natire of its insulation.
IIRR a twisted pair twisted out of enamel-insulated transformer wire
has characteristic impedance in the ball-park of 120R.
I got closer to 30 ohms last I measured a pair. Enamel is a whole lot
thinner than extruded jacketing. It's going to be even lower in a
piled-up winding due to the crowding.
The low frequency way to think of it: your leakage inductance is almost
exactly the inductance of the windings as a transmission line.
If you take a piece of twisted pair 1m long, it'll have maybe 0.5uH
inductance (measured at one end of the pair, shorting the far end, at a
frequency well below the electrical length of the line). If you wind it
up onto a form with an air core (making a bifilar solenoid, say), the
self-inductance of each winding might be a few uH, while the inductance
between wires remains the same (it's lower, if anything). Note that you
can measure this leakage two ways: terminus shorted (as a transmission
line) or secondary shorted (transformer leakage). The difference is, you
test P1-S1 and short P2-S2, or test P1-P2 and short S1-S2.
Now if you insert a permeable core, inductance goes way up (into the mH,
perhaps), and coupling coefficient likewise goes up (some fraction less
than 1.0). But leakage remains fairly constant.
Leakage depends almost entirely on winding construction. Self-inductance
depends on the windings and core. Coupling coefficient is the factor
relating the two.
(Yes, you can make a transformer that specifically depends on core
geometry, not just winding construction. An example would be two coils at
right angles, with a core snaked through each. Without the core, they
have zero mutual inductance (infinite leakage). With the core, it's
nonzero. I'm more interested in applications where you actually give a
damn about performance in the first place.
)
The important thing about transmission line transformers is to forget
about using them as transformers. Use them as transmission lines! If you
put a few loops of coax on a core and drive the shield (calling the shield
the primary, P1-P2), you can't expect any useful kind of behavior from
that, because the shield carries all sorts of crazy currents, depending on
how it's looped through, and which turns it's adjacent to, etc. If
instead you drive the transmission line from one end (P1-S1), you'll get
the same signal out (P2-S2), delayed, except the core allows you
common-mode voltage. You could flip the terminal end around (S2-P2), and
get an inverted signal!
http://www.picosecond.com/product/product.asp?prod_id=47
That's more or less what they do here. The shield necessarily does still
carry a signal (the act of flipping the terminals forces the output
voltage onto the shield anyway), but this occurs "after" the signal
propagated through, and what you do with the shield is now an open
variable -- you could loop it through a whole bunch of ferrite beads,
damping out any oscillations.
It follows that you can create any ratio by connecting transmission lines
in parallel, looping them through a core (it doesn't even matter that the
same core is used, it's just a common mode choke now!), and connecting any
desired series-parallel combination on input and output sides to set the
desired impedance and ratio.
The dirty secret of transmission line transformers is, they aren't at all
interested in reducing leakage inductance, or capacitance, or anything
like that. It's just a big common-mode choke that lets you pipe signals
from wherever to wherever else.
Tim