Impedance Balanced Outputs
Copyright 1998 by Richard A. Chinn, all rights reserved. Revised June 2000 & October 2009.
I抳e always been interested in transformerless output circuits. Over
the years, I抳e watched and played with the variations that have appeared:
the chain of inverters, the cross-coupled circuits, etc. and while each
of them solved a particular problem, none was as foolproof as a decent
transformer driven with a high-current capability amplifier.
Transformers have gotten a bad rap in the audio world. For as many problems
as they solve, they create an equal number. Sometimes, a transformer is
an elegant but simple solution to a nasty problem. Often it is unexcelled.
But with users clamoring for lower cost products, it was an easy decision
to frame the transformer as a sonic pollution dump while cutting out a
significant cost item.
Remember: marketing hype nothwithstanding, you the consumer pay for
every cost incurred by the manufacturer in making your favorite gadget.
There is no free lunch!
The issue of electronics-challenged users connecting the output to both
balanced and unbalanced inputs has always been particularly vexing. The
three commonly used solutions are the chain of inverters circuit, the cross-coupled
circuit, and ground sensing. All of these work to one degree or another,
and each has an weak point that can be troublesome.
The Chain of Inverters
The chain of inverters circuit emulates a grounded center-tapped transformer.
This is very likely the most popular circuit used today. When used with
an unbalanced input by connecting the low-side output to ground, the chain
of inverters circuit injects current (usually distorted) into your ground
system. If everything is free of pin 1 trouble, then this is probably OK.
If the unbalancing happens at the load, then this ground current is flowing
through your snake system, which often causes trouble. This unbalancing
act also severely taxes the circuitry driving the grounded output. Another
problem with this circuit is a 慺ree?6-dB of voltage gain.
The Cross-Coupled Circuit
The cross-coupled output emulates a floating transformer winding. Unlike
the transformer, this circuit demands that it be unbalanced at the source.
Heaven help you if there is a pin 1 problem designed into the unit抯 output
connector. Unbalancing at the load causes instability. The good news is
that the voltage gain is constant, balanced or unbalanced. The Analog Devices
SSM2142 and Burr-Brown DRV134 IC parts use this circuit.
The ground sensing circuit attempts to make whatever difference in potential
between the near equipment ground and the far equipment ground into a common-mode
signal. To one degree or another, this usually works. It is immune to having
its low side output shorted to ground. Sometimes it makes matters worse.
DBX, TAPCO and Dolby have all used this circuit at one time or another.
Let's Blame the Manufacturers
All of this is exacerbated by most manufacturer抯 omission of any discussion
of output stage topology in their manuals. Of course, if they were considerate
enough to include a schematic (not often), then anyone with a bit of circuit
smarts can see this out at a glance. Some manufacturers now include a discussion
of balanced and unbalanced interface issues along with specific recommendations
for their unit. I believe that they should also describe their unit抯 input
and output circuit topologies.
As an equipment designer, I抳e agonized over this problem for some time
now. One solution is to simply provide two connectors: one marked balanced,
and one marked unbalanced. This works, and doesn抰 really take a rocket
scientist to figure out (good when the user refuses to read the g!@**!
manual. On the other hand, it requires two connectors along with more PCB
space and panel space (both cost money, and panel space may be limited).
I抳e also gone so far as to insist on using TRS connectors for the unbalanced
outputs, wiring the ring contact to circuit ground. This way, connecting
the gear using a TRS cable into a balanced input works just fine. This
is a win in more ways than one, both for the user and for the manufacturer.
The manufacturer gets to buy one less different part and the user gets
a jack that accepts either a TRS or a TS plug and connects each one correctly.
The people at Mackie Designs have played with a scheme that they called
"Impedance Balanced" for phone jack outputs. At first, I gasped and sputtered.
But after thinking about the problem, I抳e come to embrace this as the
best solution, especially when you don抰 have dual output connectors (if
product cost is a factor, then every nickle and dime counts, and excess
connectors are a great place to shave costs).
Simply stated, you use a TRS jack for the connector, and you connect
the ring contact to ground through the same resistance as used for the
build-out resistor on the tip contact. From the line抯 standpoint, the
circuit is balanced, driven from a grounded center-tap source. Only half
of the circuit is driven from an audio standpoint. It doesn抰 matter what
sort of plug the user inserts, either way is optimum for that case. For
the unbalanced case, no ground current flows.
If you're building this into a piece of gear, then take the time to try to figure
out what the value of the buildout resistor is from the signal source, then match that
value from the ring side to ground. If there is a coupling capacitor, then add another
of equal value in series with the resistor going to the ring contact.
Even if you can't build this into the gear, you can take advantage of the balanced input
by running the balanced input all the way to the source, and then connecting the hot side to signal
and the cold side and shield to ground. While not as good as impedance balancing,
this is far superior to treating the whole mess as an unbalanced interface.
From the balanced input抯 standpoint, there are equal impedances from
both sides of the line to ground. This results in the best CMR performance,
especially if close-tolerance resistors are used at the output side.
From an induced noise standpoint (the kind of noise that a balanced input
is supposed to reject) this is what counts. There is no level difference
if used with a balanced or unbalanced input. The circuit noise is actually
3 dB less than either of the other two balanced line drivers. This is
way better than unbalancing the input at the input jack and handling
the interface as unbalanced output driving unbalanced input.
The really important issue is that both sides of the line see the same
impedance to ground. It is far less important that the line be driven symetrically
(i.e. both sides antiphase).
Lest you think that this is some sort of half-baked solution, take note
that both AKG and Neumann use this circuit for their transformerless microphones.
If it抯 good enough for them at microphone signal levels, I抦 more than
comfortable with it at line level.
doug self has a more math intensive explanation on his