Antenna Tuner Baluns

Tuner Baluns on Input

 Sleeve baluns   
Balun and Transformer Core Selection       
4:1 Balun
Analysis   
Verticals and Baluns   

    Common
mode current   
Common-Mode Noise   
toroid balun winding   
Steel wool balun

10/2/2012 added data differential loss

A common misconception is
placing a balun on
the input of a tuner
causes the balun to
work better. The
thought or claim is the balun
operates with a
matched impedance,
and that reduces
balun losses. It
also is thought or claimed
moving the balun
to the tuner input improves balance.

Any article that claims placing a current
balun at an unbalanced
tuner input helps balance,
or generally helps the balun work better, unfortunately
is mistaken. Moving the current or choke balun to the input of an unbalanced
network makes balance much worse on higher bands, and does not help lower bands.

Let’s take a close look
at what actually happens when we move the balun to the tuner input.

Spice Models

Ideal
Case Tuner with
Balun on Input

The following
model has a near
perfect balun on the input
of a perfect unbalanced
network. In the
perfect case below
ill-effects caused
by floating the
network, such as
unbalanced
coupling to
ground from the
network components, is
ignored. Virtual components lack stay  coupling to the chassis and cabinet.
This is a
much better tuner than
any of us can build. 

The load
is perfectly
balanced with 1000
ohms total
impedance.  In this
example, balun common-mode impedance is purely
dissipative. 50K
ohms would be a VERY
good balun, with an impedance
nearly impossible to
obtain on any frequency. A practical balun would be much worse, but let’s look at an
extremely good case.

In the circuit below, at point “A” (R1),
power dissipation is
indicated by the
green trace. Point “B”
or R2’s power
dissipation is
represented by the
yellow trace. All
voltages and power
levels are
instantaneous peaks.
The capacitors and
inductors are
assumed lossless.
Applied “power” can
be considered to be
about 30 watts.

Note: Resistor R3
represents
represents the
common mode
impedance of the
balun. The windings
in T2 have unity
mutual coupling,
similar to a real
bifilar balun. This
is a simple but
reasonably accurate
way to represent the
common mode losses
of our ideal 1:1 current
balun.

In the above case
we see the load is
very well balanced.
The green and yellow
peaks are nearly
identical. The load
receives well
balanced power.
Balun heating is 1%
of R2’s power.

String of Beads
Balun on Tuner Input

Let’s see what
happens when a more
typical “very long string of
beads balun” is
used. In this case we will ignore balun flaws and assume ideal resistance, with
zero stray or shunt capacitance: 

In this case power
in the upper side of
the balanced load,
represented by R1,
increases. There is
also a shortfall of
power in the lower
side of the load.
Power input slightly
increases (I didn’t
readjust the network
for the changed
impedance) to 32.5
watts 

R1 = 18 watts   
R2 = 11.5 watts 
We have 6.5 watts of
load unbalance. The
balun dissipates
about 3 watts.

Balun on Tuner
Output

Now let’s compare
the system above
with a system having
the same string-of-beads balun on the
tuner output:   

Notice all power
levels are
unchanged!

• Unbalance
  in the load is the
  same
• Balun
  heating is the
  same

We lost
absolutely nothing
by moving the balun
to the output!

With the balun on
the input we have
the following peak
common mode voltage:

We have about 75
volts peak across
the 2K ohms of balun
common-mode
impedance, or about
3 watts dissipation.
This is about 10% of
applied power. With
1500 watts the balun
would be dissipating
150 watts. The beads
would be too hot to
touch in a matter of
seconds.

Now let’s examine
heating caused by
common mode voltage
with the balun on
the output.

Notice we have
exactly the same
balun heating power.
Nothing changed. The
balun core is under
exactly the same
electrical stress.
Nothing was gained
by relocating the
balun to the input.
Nothing was lost by
moving it to the
output! The same
core size is
required, balance is
the same, and
heating is the same.

In the real world, we would find balance WORSE with the balun on the tuner
input. This is because a typical group of medium sized matching components would
have about 50 PF or more capacitance to the chassis on forty meters, with a very
large portion of capacitance on the stator side of the tuning capacitor and the
“ground” side of any inductor. In every tuner I have tested with an unbalanced
network and a balun on the input, balance has been very poor on higher
frequencies.

Real Component
Test

Let’s look
at real components
and see how they
work. First let’s
measure the common
mode impedance of
the bead balun I’m
going to use:

This is the tuner
configuration. In
this case it was
setup for a
floating-L network
with balun on the
input:

The ground side
of the floating L
measured 10mV at the
load:

The “hot side” of
the floating L
measured 185mV:

This is
worse than the model
because the beads do
not have the 2000
ohms impedance the
model assumed. The beads
only have a few
hundred ohms
impedance. The model also does not include significant network-to-chassis stray
impedances that are present, which unbalance the output terminals.

The large amount of unbalanced coupling from network components to chassis is
commonly overlooked. Components are not physically or electrically
identical when viewed from each balanced load terminal to “ground”. Impedance to ground is
not nearly as high as we might expect, and impedance to ground varies greatly with actual
component settings and operating frequency.

More Accurate Model Representation

In a real device with real components, stray impedances distributed through
the system create significant problems. For example, above 40 meters there is
significant change in balun impedance from the addition of ceramic-insulated
feedthrough terminals, wire connections, metal cases, and even dielectrics. We
might eliminate the metal case to reduce capacitance, but without a metal
enclosure everything around the balun affects common mode isolation and balance.
Many devices test very well on a bench, but do not hold the expected performance
in the real environment.

Based on my considerable experience, which includes thousands of hours
evaluating baluns and tuners, it is foolish to go overboard with impedance.
Also, some of the poorest systems appear to offer the best performance. A balun
operating at high power and remaining perfectly cold often indicates a balun
inadequate for the job.

The circuit below, while still incomplete, more closely represents a real
tuner.

CS1 through LS1 are shunt impedances.
They are capacitive at root, but combine with series inductive reactances and
losses. Overall, they will always be partly resistive, and may be
capacitive through inductive in actual net effect on the system.

Making matters worse, values of everything significantly change with
operating frequency, setting of L1, and capacitor C1 and C2 settings.  L1
is a single inductor.

Since parasitic inductive and capacitive reactances are in series, the net
reactance is reduced and can sometimes reach very low values!  

Looking at L1, we find various areas of L1 are in series to balanced output
terminals B1 and B2. The reactance of these areas varies as  L1 is
adjusted. The effect of this is the stray path through LS1 to the tuner
enclosure varies in series L, C, and R as the roller is adjusted. If we watch
terminals B1 or B2 to ground, looking at common mode impedance, we find both the
ratio and absolute impedances vary. Lower inductance settings push the lowest
impedance frequency of B2 lower, because L1b and c get larger. L1a is always in
series with B1, so the circuit is asymmetrical as viewed from antenna terminals
B1 and B2.

Further complicating matters and aggravating symmetry problems, C1 and C2
have varying CS1 and CS2 as the capacitors are rotated. Since the stray of CS1
and CS2 to chassis is normally capacitive across HF into VHF, it parallel or
series tunes L1’s sections in varying amounts. The direction and amount varies
with frequency, and is never symmetrical at B1 and B2.

When viewed on a Smith Chart, terminals B1 and B2 are typically a series of
circles spiraling from reasonable values of expected shunt capacitance on 160
meters, circling through inductive and capacitive swings, down to a fairly low
and unequal shunt impedance on 30 MHz.   

Casual designers look at the tuner and assume B1 and B2 (balanced terminals)
represent some impedance equal or higher than BALz, but that is never the case
unless through chance the equivalent effect of all strays parallel tunes BALz to
a high parallel resonant impedance. There probably isn’t a worse system for consistency and reliability in common
mode suppression as obtained when a balun is moved to the input of an unbalanced
network, especially with physically large components.

The Roller Alone

Shunt capacitance to ground is
problematic with baluns, and large components in networks are much worse. This
problem is aggravated by series inductive reactance in the network. For example, here is the roller inductor frame-to-chassis impedance of a
relatively small roller inductor. This roller has adequate inductance to cover
160-meters, and is spaced 2 inches above a large flat copper sheet. The roller
is set near a typical position for 40 meter operation:

At 10 MHz and 25 MHz, roller frame to chassis impedance goes through a
minimum. Inductor terminal impedance, and terminal impedance variations with
frequency, are not symmetrical on the “cold” (roller wheel and frame terminal)
and “hot” (capacitor junction terminal)
side of the inductor. Resonant points and terminal impedances to the chassis
vary greatly with roller wheel position, making it impossible to
balance matching network terminal impedance to chassis.

A balun on
the input of a
floating unbalanced
network is a waste
of time. The system models this way, and the system also
measures this way. It is just a very bad idea for system balance. 

The Balun (Common Mode Choke) Loss Argument

Some have proposed balun conductor heating due to SWR on the balun’s
transmission line is a major concern. Several important things are missed in the
linked analysis.

The idea differential-mode (transmission line mode) heat is the dominant problem is, at best,
incomplete. Most people with real-world balun experience would disagree with
claims differential mode heating is a dominant design issue. Unless operated at
very low load impedances, the dominant heat mechanism in any normal current balun is
from common-mode (choking) core excitation. Differential mode, or transmission
line mode heating problems pale in comparison to common mode heating problems.
The example linked uses a terrible balun design for many reasons. Let’s look at
the 50-ohm transmission line used in the above linked analysis.

When we place a balun or common mode choke directly at a matching network’s
input or output terminals, the balun or choke’s differential mode impedance (transmission
line impedance used inside the balun) does not need to match the system.
Provided we do not transform to difficult impedances at higher frequencies where
the transmission line might be of electrically significant length, standing waves
in the balun can be compensated though simple
matching network readjustment. As a matter of fact, 50 ohms is a very poor choice of balun transmission
line impedance for this application. We ideally want differential impedance
(transmission line impedance) of a tuner output balun
to be the geometric mean of the highest and lowest expected impedances
seen on the
highest bands. Designing at the geometric mean results in minimal
differential mode impedance mismatches and
minimizes unwanted impedance transformations.

Let’s assume a practical matching network has a useful ten meter impedance range of 15-2000 ohms.
Further, let’s assume antenna systems fall with equal likelihood anywhere in a
15-2000 ohm range. In this case, the designer should choose the geometric mean of the
highest and lowest impedance. This would be (sqrt of 15*2000) a ~173 ohm balun differential mode
(transmission line) impedance. A well-planned design would use the smallest
physical size twisted pair that safely handles maximum expected differential mode
average current and highest instantaneous peak voltage. At 15-2000 ohm
impedance range at 1500 watts, current limits would dictate conductors sized to
carry over 10-amperes intermittent current without overheating failure, and
2450 peak volts of dielectric strength. Since dielectric failures are generally
instantaneous, it is important to include extra margin. Because of transmission
line effects, peak voltage can exceed maximum rated network component voltages,
so we want to allow for a certain maximum load impedance and peak power plus
headroom. Voltage requirements are easy to meet, most E or EE Teflon wire withstands
several thousand volts.

Once we determine impedance, we should consider standing waves inside the balun transmission line on
higher bands.  With ten feet of 150 ohm twisted pair, and with an SWR as
high as 13.4 : 1 (2000/150) at the high voltage end, we could have 11.25 ohms at the
network. This would be  11.55 amperes RMS current. This would call for a
fairly heavy wire. Using EE Teflon insulated stranded wire, #16 gauge would
handle 1500 watts in normal ICAS service. Note this also keeps impedance inside
the network’s range.

With proper balun design, standing wave ratios inside
the balun are reasonable. SWR will not come close to 60:1. By far, the largest balun heating effect would be from core excitation due to
common mode current flowing through the balun windings (I^2*R core heat), or if
we prefer, heat from voltage impressed across the windings from end-to-end
(E^2/R core heat). With a typically obtainable internal 2500 ohm j0 CM impedance at
optimum frequency in a very good balun or CM choke, core heat from 1500 watts
into a 2000 ohm balanced load would be about 300 watts. This does not get
any better if the balun is relocated to the network input. Looking more than superficially, the “lower loss
justification” quickly goes away.

We have not considered balance. If we
want the tuner’s output terminals to be balanced as closely as practical
(it is almost impossible to closely balance a high impedance line at upper HF),
we would quickly discard any notion of placing a balun at the network input.

When does
moving the balun
help?  

Moving the balun
helps performance
only under very
specific conditions.

1.) When the
network is a double-balanced FLOATING
network that is
perfectly
symmetrical

In this case, it works because we almost don’t need the balun. Absent any common
mode current fed
back from the
antenna, the move
will reduce
excitation of the
balun core almost
completely. In this
case, the balun
actually has very
little use at all!
It can be eliminated
in most cases with
very little effect
on system balance.
With the balun on the input of a balanced tuner, the
tuner becomes a
balanced voltage
source.

2.) When the
network is a double-balanced network
that is perfectly
symmetrical and
grounded in the
center

In this case a
modest balun is
required, but the
tuner becomes a very
“stiff” split voltage
source.

It’s actually
just as difficult to
build the best kind
of tuner, a
tuner that provides a balanced current source with high common mode isolation,
by using a double network. If we spent only a fraction of the additional
component expense required to build a “balanced tuner” and used the money to
build a good
current balun, we
could have a better-balanced tuner in
much smaller space!
Instead of working
towards a good
current balun,
some designers have
thrown money into a
system that creates more new problems
then it cures old
problems.

Balanced tuners
are great when the
load is closely
balanced. They are
not so good feeding
Windom, Zepp, or
other antennas operating in the
nether world between
perfectly balanced
and perfectly
unbalanced. They
also generally, for the same
dollar
investment, will not
outperform a
conventional unbalanced network
with a properly designed balun on
the output.

You can see actual load balance measurements, including phase,  on this page.

Ironically, moving the balun to the input only works when the balun
is not needed!

Since 9/13/2006    Contents copyright
W8JI 2006-2012