Balanced ATUs and common mode current

This Feb 2012 article has been copied by request from my web site which is no longer online. The article may contain links to articles on that site and which are no longer available.

Many designs have a ‘balanced output’ or an option of a ‘balanced output’, but what does that mean, and are they effective in minimising common mode current in an antenna feed line?

ATUs achieve ‘balanced output’ in one of several ways, the common ones are:

  • a grounded impedance transformation network followed by an internal voltage balun;
  • a grounded impedance transformation network followed by an internal current balun;
  • a current balun followed by a symmetric impedance transformation network that may or may not be directly grounded at its centre;
  • a link coupled ATU where the output circuit is symmetric and may or may not be directly grounded at its centre.

Much has been written about the merits of one approach or another, mostly qualitative and often subjective, but there is little in the way of quantitative analysis of the impedance that the ATU offers to common mode current.

The less than scientific approach to the topic leads to the ‘old is new again’ and ‘going back to the old and proven designs’.

Well, they are old, but were they proven?

Over the years, there is a stream of articles that proposes designs that are more suited to balanced antenna systems, but none that I have seen provide quantitative support for their designs. They certainly provide fodder for magazines, and the article this month in QST (Pituch 2012) with the headline “For matching balanced antenna systems, the classic link coupled tuner is hard to beat” is yet another example, nothing too original, mostly “what is old is new again”.


To understand the problem, we first need to visit antenna systems using two wire open feed line and the meaning / nature of ‘balance’.

If we want the feed line to perform the function of transferring energy from transmitter to the nominal radiator without contributing to radiation itself, then we want that at all points along the feed line, the current in one conductor is equal in magnitude and opposite in phase to the current in the other conductor so that to the maximum extent possible, their contribution to external fields cancels.

This is not always the goal, but for most applications where a general purpose HF balanced ATU is used, this is the most rational goal.

Common mode current can often be measured readily, see Measuring common mode current (Duffy 2011b).

Common mode impedance is the impedance offered to the common mode component of line transmission line current, and is the most important characteristic of a balun intended to drive equal but opposite currents.

(Witt 1995) and (Witt 1995b) describes tests on several ATUs including a balance test. The balance test will give a perfect score of 100 in Witt’s metric when a tuner drives equal but opposite voltages to his symmetric loads, though it may have very low common mode impedance. Witt’s test is not a good indication of the tendency to drive equal but opposite currents into an asymmetric load. Despite Witt’s balance test being quite worthless, the reports are often cited.

Antenna symmetry

If a radiator is perfectly symmetric in every respect, radiator dimensional symmetry, parallel to ground, orthogonal feed line routing, homogenous ground (includes even moisture content), no buildings or vegetation nearby etc, the voltages on the feed point terminals will also be symmetric if driven by equal / opposite currents. So, driving the feed line with equal / opposite voltages will results in equal / opposite currents and the system is balanced in every way, feed line radiation will be minimal.

But, antennas are not necessarily or even usually perfectly symmetric, the question is more one of how imperfect.

Common mode current counter measures

There are a number of counter measures, and none are guaranteed to eliminate common mode current… well apart from disabling the transmitter! Reduction of common mode current to a sufficiently low level that it is not troublesome is the best that can usually be achieved.

A very common counter measure is to include a substantial impedance in the common mode current path at one or more places. Such a device is often termed a common mode choke, or an isolator, but it is also simply a 1:1 current balun. The meaning of substantial warrants discussion, there are plenty of Rules of Thumb, some have been peddled for many decades, but most are not supported by quantitative analysis of antenna system scenarios.

(Duffy 2010) shows a method of applying common mode impedance to an NEC model of an antenna system to explore the effect on common mode current.

Wouldn’t it be good if the ATU offered a high impedance to common mode current.

Balanced ATU common mode impedance

Ideal floating link coupler

A link coupled tuner where the link coupling is very loose (ie very low percentage of the primary coil flux links the secondary coil), and that is physically symmetric with respect to grounded conductors, and with very little stray capacitance to ground will offer a very high common mode impedance. Essentially, current that flows in one output terminal must be accompanies be an equal but opposite current in the other, there is no current path to ground through direct ground connections anywhere in the output network, or via stray capacitance.

In the days of push pull valve PAs with symmetric tank circuits and a loosely coupled symmetric link to the feed line, current balance and common mode impedance might have been very good, but it should not be assumed that link coupling equates to high common mode impedance, the implementation matters.

Fig 0: Output circuit of the original Z-match
Fig 0: Output circuit of the original Z-match

Fig 0 above shows the output circuit of the original Z-match from (King 1955). The circuit was implemented in the KW E-ZEE Match. The unbalanced input is at the terminal at the top left, and J4 is used for 80 and 40m output, J3 used for 20 to 10m output. Note that L1 is grounded at one end, and L3 has almost as many turns as L1 (6.5t vs 7.75t) and is wound a little larger in diameter and over L1, so fairly tightly coupled. This is not an electrically symmetric isolated link, and can be expected to have reduced common mode impedance.

The more recent single coil Z-match has the same issue of a tank coil that is grounded at one end, and the link cannot be electrically symmetric and there has been online discussion of the imbalance though some tests are not sound (eg (Butler 1996) ignores phase in his test).

Link coupled output circuit with a current path to ground

When the link drives a circuit with a current path to ground, common mode impedance is compromised by that current path. The extent of the compromise depends on the circuit configuration, values, frequency etc.

Consider a simple link that is tuned by a 50+50pF lossless capacitor with the centre grounded. If the link is assumed to have infinite common mode impedance, the common mode impedance of the link and tuning capacitor is around 0-j113Ω at 14MHz, a very low common mode impedance that is unlikely to significantly reduce the prospect of common mode current, indeed being capacitive, it will offset some of the effect of a cascaded effective common mode choke should one be used to address the problem.

Some link coupled output circuits are rather complicated, and the common mode impedance will be very dependent on the settings for a particular load, frequency etc, but if some point in the output network is grounded, common mode impedance will usually be much poorer than the loosely coupled link alone.

Fig 1: Output circuit of Johnson Matchbox

Fig 1 above shows the output circuit of the popular 275Watt Johnson Matchbox, the Kilowatt Matchbox is similar. Note the ground connections to the variable capacitors C1 and C2. Grounding the output network in this cases reduces common mode impedance.

(Cebik 2007) explains link coupled tuners and starts of with an isolated output circuit, but develops the circuit to ground the output circuit without explanation of why he has done so, or discussion of the detrimental affect on common mode impedance (which itself is not mentioned). His discussion of the importance of symmetric capacitors with grounded centres to balance reveals that his design objective is voltage balance at the expense of current balance: Capacitor construction is equally important in maintaining circuit balance with respect to ground. Split-stator capacitors provide an inherently balanced structure, with roughly equal influences on both sides of the circuit from stray capacitance to a metal case or other metallic objects in the circuit.

Fig 2: Pituch’s circuit

Fig 2 above (Pituch 2012) circuit which derives from Cebik. It has a directly grounded output circuit, reducing common mode impedance.

Symmetric balanced tuners and stray capacitance

There are a number of ATU designs that claim superior performance because the ATU network is symmetric / balanced, and the transition from the unbalanced transmitter output to balanced condition is performed in a balun between the ATU input and the impedance transformation network.

Voltage balun at input

If the input balun is a voltage balun, the whole ATU will probably have low common mode impedance at the output terminals.

Current balun at input and grounded output network

Some designs ground some point in the impedance transformation network and so, even if they used an effective current balun at the input, common mode impedance will be reduced by the grounded network. Though grounding the output network was common in the link coupled generation of ATUs, direct grounding of the output network in modern designs using a current balun at the input is less common.

Current balun at input and floating output network

There is at least one design (Measures 2003) that uses an air cored multi turn coax choke for the input balun. This type of choke typically exhibits a very high common mode impedance at its resonant frequency, and very low common mode impedance everywhere else, they are typically narrowband chokes. No common mode impedance measurements or predictions are offered by Measures.

Current commercial products employing a ferrite cored current balun at the input, and a floating symmetric impedance network include Palstar BT1500A, MFJ-974H, MFJ-976. None of those products specific common mode impedance in the form of a graph of Rcm and Xcm vs frequency. Note that for these topologies, common mode impedance will be to some extent dependent on the tuner settings, so providing a simple meaningful specification will be challenging.

Even if they use an effective current balun on the input, and floating output network, housing the components within a close fitting conductive enclosure will provide a current path to ground through the stray capacitance that exists (as demonstrated earlier), and although not as bad as grounding a point in the network, the stray capacitance path degrades the common mode impedance achieved with the input current balun.

Measurement of the common mode impedance reveals the design achievement, but designers and manufacturers rarely publish such data.


Fig 3:

Fig 3 is from (Danzer 2004). Danzer discusses methods of feeding open wire feed line without mentioning current balance, or common mode current.

He dredges up a 1964 quote to criticise the configuration in Fig 3(A) as follows.

Strangely enough, the balun problem has been known “forever,” and until only recently it has been ignored. In the 10th edition (1964) of The ARRL Antenna Book, the following was printed on this topic:

The principles on which balun coils operate should make it obvious that the s.w.r. on the transmission line to the antenna must be close to 1 to 1. If it is not, the input impedance of each bifilar winding will depend on its electrical characteristics and the input impedance of the main transmission line…and the transformation ratio likewise will vary over wide limits.

Translated, this means that if the balun is not operated under matched conditions, it does not operate as the simple unbalanced to balun device you might imagine. Moreover, from a practical point of view, the balun core will get hot. This means that some of the power you thought you were sending to the antenna is actually going to raise the temperature of your tuner and your shack. There are easier ways to heat the radio room!

This ARRL positions is more than fourty years old, it is dated and quite flawed but it has passed through QST’s review process in 2004.

A current balun with high common mode impedance deployed as in (A) will be effective in reducing common mode current. Contrary to the quoted statements, it does not have to be operated with VSWR=1 to be effective in reducing common mode current, though it is true that it contains a transmission line section that will transform impedance. The latter is not usually an issue as open wire lines are usually operated at high VSWR and the purpose of the ATU is to perform the necessary impedance transformation.

As far as heating goes, a Guanella 1:1 current balun develops flux due to the common mode (or imbalance) current, and that heats the core, more so for lossy cores. The solution is to use one or more current baluns with sufficiently high common mode impedance to reduce the current sufficiently that little power is lost in the baluns. Of course, addressing the antenna system asymmetry that drives the unbalanced current is a more direct solution.

Danzer’s ‘high efficiency’ option at (B) is based on (Straw 2007, p25.15). Straw’s design is hardly symmetric, and he gives no measurements or predictions of its common mode impedance or ability to minimise common mode current. It seems that Danzer’s recommendation of (B) over (A) is based entirely on the ARRL’s quote which is seriously flawed, and there is no measurement data to backup the claims.


(Straw 2007. p25.15) describes an ATU that is a current balun followed by an asymmetric T match that floats above the chassis, in fact the T match is built on a sub chassis that is bonded to the T point, and this sub chassis is capacitively coupled to the exterior enclosure. It is this project that is the inspiration for Danzer’s Fig 3(B).

The article does not mention current balance, common mode impedance, or offer measurements of its performance in reducing common mode current.

Straw discusses one measure he took to preserve balance

In our unit, a piece of RG-213 coax is used to connect the output coaxial socket (in parallel with the “hot” insulated feedthrough insulator) to S1D common. This adds approximately 15 pF fixed capacity to ground. An equal length of RG-213 is used at the “cold” feedthrough insulator so that the circuit remains balanced to ground when used with balanced transmission lines.

Assuming the equivalence as valid, let us explore the effect of that 15pF capacitance on common mode current for an assymetric load of 3000+j0Ω. (An extreme load, but Straw has used 3000Ω in some of his design justification (Straw 2011).)

Fig 4: Straw’s output circuit with extreme assymetric load


Fig 4 above shows the equivalent circuit driven by a balanced current source Is, and connected to a three terminal extreme asymmetric load. Let us assume that the three terminal load equivalent circuit is that presented to the ATU (ie adjacent to the ATU).

Fig 5: Effect of Straw's capacitance on common mode current
Fig 5: Effect of Straw’s capacitance on common mode current


Fig 5 above shows that for a perfect current source Is feeding the network, the current balance is destroyed by the effects of the shunt capacitance. The model does not include all shunt capacitance that would exist in a practical implementation, just the coax element discussed by Straw, and so underestimates the effect.

Straw does raise the prospect of voltage breakdown in baluns on the load side of an ATU.

Include a balun operating within its design impedances. Often, a balun is added to the output of a tuner. If it is designed as a 4:1 unit, it expects to see 200 Ω on its output. Connect it to ladder line and let it see a 1000-Ω load, and spectacular arcing can occur even at moderate (100 W) power levels.

Again the notion is that baluns only ‘work’ at some design impedance, and the anchoring to 4:1 voltage baluns in old designs. A 1:1 Guanella current balun with high common mode impedance can be effective in reducing common mode current irrespective of the differential impedance at that point. To be used on the load side of an ATU, it does need to be designed with high voltage withstand.


(Hallas 2004) article which is labeled as a product review leverages Danzer’s article and introduces the problems of tuners with the usual 4:1 balun. The usual 4:1 balun is a voltage balun, and as will be discussed later, will achieve current balance only on symmetric loads, and antenna systems are not necessarily symmetric loads.

The article refers to one of the commercial tuners as using the insulated unbalanced scheme, Fig 3(B) in Danzer’s article. This is a DC view of the circuit, the thought that the network can be DC isolated from the exterior box and that assures RF symmetry is wrong, it does not address the reality of stray capacitance. The article contains no mention of current balance or common mode current, has no measurements of common mode impedance or measurements of common mode current.

Voltage balun at the output

Of the methods commonly used in ATU designs to achieve ‘balanced output’, all of those that depend on a voltage balun at the output have very low common mode impedance and so are less effective in assuring current balance on asymmetric loads.

Most T match ATUs that have a balanced output option use a voltage balun. Examples of commercial products include the MFJ-949E.

Current balun at the output and stray capacitance

Some T match ATUs have an internal 1:1 current balun to provide an optional balanced output. Examples of commercial products include Ameritron ATR-30.

Just as stray capacitance to ground degrades the common mode impedance of a symmetric ATU, it does also for a current balun. A common construction technique is to clamp a toroidal current balun against an conductive panel.

Fig 6 above shows the current balun in a popular ATU, the Ameritron ATR-30, clamped against the rear conductive panel.

Fig 7: ATR-30 common mode impedance


Fig 7 above shows the common mode impedance of the current balun in Fig 4. It is unusual for an ATU to incorporate a current balun, and this one has very high common mode impedance around 4.8MHz where it exhibits self resonance, and lower common mode impedance elsewhere. It shows all the hallmarks of design using low permeability low loss ferrite cores (similar to #61 mix), for a narrow self resonance and lower choking impedance elsewhere. The combination of choke inductance and self and stray capacitance cause the resonance quite low in frequency (for a 1.8 to 30MHz ATU), and the low loss core material contributes to the lack of choking impedance well above resonance.

Measuring common mode current as described at Measuring common mode current,  at the ATU output on a real antenna, a 40m dipole, the magnitudes of currents in each wire are indistinguishable at 850mA each, and the magnitude common mode current (ie both wires passing through the current probe) is below the lowest instrument graduation at 10mA, so the common mode component in each conductor is about 0.5% of the differential mode current. (At such low readings, it is not possible to reliably estimate the phase of common mode relative to the differential mode current using the amplitudes of each.)

Fig 8 VK1OD HF balun common mode impedance

Fig 8 above shows the choking impedance for a ‘home brew’ design (Duffy 2011) housed in a non conductive enclosure, using a higher permeability lossier core material (F14 mix, similar to K and 52 mixes).

Fig 9: VK1OD HF balun with additional 10pF self capacitance common mode impedance

Fig 9 above shows the choking impedance for a the same balun as in Fig 6 with an additional 10pF of capacitance to ground. The small additional self capacitance has moved self resonance down to 4.2MHz, and has reduced common mode impedance at 30MHz by 64%.

Note the similarity to Fig 5. Adding a further couple of pF of capacitance to the configuration in Fig 6 produces almost the same result as depicted in Fig 7.

Stray capacitance compromises common mode impedance.

The common mode impedance of a Guanella 4:1 current balun is usually less than that of the constituent 1:1 baluns. 4:1 baluns provide impedance transformation, but are usually poorer at reduction of common mode current.

There is good argument to house a high common mode impedance 1:1 current balun in a non-conductive ventilated enclosure, external to the ATU, connected by a very short length of low loss coax. Reduction of stray capacitance reduces common mode impedance degradation at higher frequencies where the reactance of the stray capacitance becomes more significant.


  • There is a quest in ham radio for an ATU optimised for use of open wire feed line.
  • Common mode impedance is a key metric for assessing the performance of an ATU as part of an antenna system.
  • It is their ability to minimise common mode current that differentiates balanced ATUs from other ATUs.
  • Descriptions of designs in the ham literature do not usually define the ‘balance’ objective, and do not usually quantify performance, much less measure or predict common mode impedance.
  • Link coupling is not inherently current balanced, the implementation has bearing on the common mode impedance.
  • Designs commonly incorporate elements that compromise common mode performance, often delivering very low common mode performance that is little help in reducing common mode current.
  • Older designs aren’t magic, we just didn’t understand their shortcomings at the time.
  • A unbalanced T match followed by a 1:1 Guanella (current) balun in an external non conductive enclosure and having high choking impedance, very short coax connection to the ATU, and high voltage withstand is capable of excellent performance in a ‘balanced ATU’ role for general purpose HF application.

Link / References


Version Date Description
1.01 22/02/2012 Initial
1.02 01/07/2015  Copied to owenduffy/net/blog.