[LAU] turning a consumer soundcard into "prosumer" w/ quasi-balanced outs

[Niels Mayer]

On Sat, Jun 5, 2010 at 2:48 PM,  <fons at kokkinizita.net> wrote:
>> Despite popular belief, this is not necessary for noise rejection. As
>> long as the impedances are balanced, noise will couple equally into
>> the two wires (and be rejected by a differential amplifier),
>> regardless of the signal that is present on them.[1][2] A simple
>> method of driving a balanced line is to inject the signal into the
>> "hot" wire through a known source impedance, and connect the "cold"
>> wire to ground through an identical impedance. Due to common
>> misconceptions about differential signalling, this is often referred
>> to as a quasi-balanced or impedance-balanced output, though it is, in
>> fact, fully balanced and will reject common-mode interference.
>
> Common mode interference will be rejected provided the input at
> the other end is truly differential. Many inputs on consumer or
> pro-sumer type of equipment are not, even if they are presented
> as such.

I'm thinking of the scenario of driving a mixer with balanced inputs,
or active speakers with balanced inputs. I believe mixers by Mackie
and other reputable companies have true differential inputs.

> But calling such an output a 'fully balanced output' is
> stretching the truth - describing it as such in a spec
> or sales literature would be a plain lie.

You mean like M-audio?
http://forums.m-audio.com/showthread.php?869-delta-1010-balanced-jack-query
... I'm sure http://behringer.com does it too. What I'm wondering
about is whether their balanced inputs are also faux? Now that would
truly be a lie because you can't fake a differential input, except
using a tranformer ( and http://www.jensen-transformers.com/ are more
expensive than adding an op amp.)

For quasibalanced OUTPUTS, however, I disagree that it is a lie. It
really is a balanced output. If you do a
Einstein-relativity-gedankenexperiment except instead of riding an
elevator you're riding the balanced cable, it is clear that without a
reference point (which is what a true differential input would provide
-- a non-reference point) this output is exactly like  having both (+)
and (-) side of the differential pair actively driven; the only
difference is that you can't attain the "pro" output levels with such
an output, which is what they mean by the "gain reach issue" in the
application-note from Jensen Transformers, Inc (which I consider to be
a reputable pro-audio company, and also note that none of their
suggestions I posted here involved using their expensive transformers
:-) )

Furthermore, unless the differential pair output opamps are
well-matched, which is unlikely for mass-production prosumer cards,
you won't have a balanced output, nor will it necessarily have good
audio characteristics nor CMRR.  In that case, you might actually be
better off only having to worry about a single output driver opamp and
it's regular-tolerance components wiith a quasi-balanced out; rather
than needing to have a pairs  of them precisely matched as needed for
a true balanced out. With a pair of output stages, you double the
noise-floor, and distortion characteristics of the output stage.
Therefore quasi-balanced outs might actually be a blessing for
prosumer cards, where they likely skimp on these output components, so
they'll only sound half as bad as they could have with a true
differential pair. :-)

http://en.wikipedia.org/wiki/Balanced_audio#Differential_signalling
continues by outlining the features missed with a quasi-differential
output:
..........................
However, there are some benefits to driving the line with a fully
differential output:
The electromagnetic field around a differential line is ideally zero,
which reduces crosstalk into adjacent cables.
Though the signal level would not be changed due to nominal level
standardization, the maximum output from the differential drivers is
twice as much, giving 6 dB extra headroom.[1] (if the amplifiers are
identical, though, their output noise sums to 3 dB more than a single
amplifier, decreasing dynamic range).
Noise that is correlated between the two amps (from imperfect power
supply rejection, for instance), would be cancelled out.
At higher frequencies, the output impedance of the output amplifier
can change, resulting in a small imbalance. When driven in
differential mode by two identical amplifiers, this impedance change
will be the same for both lines, and thus cancelled out.[1]
Differential drivers are also more forgiving of incorrectly wired
adapters or equipment that unbalances the signal by shorting pin 2.[1]
..................

> There are many variations for both 'balanced' inputs and
> outputs. For outputs there is at least:
>
> 1) Impedance matched, as described by the OP.
>
> All the following must also be impedance matched, exept
> the last for which the concept doesn't make sense.
>
> 2) Providing signal on pin 2 and inverted signal on pin 3,
>   both signals are independent. Connecting such an ouput
>   to a single-ended input will short-circuit one output
>   (which may or may not survive, and may or may not
>   produce nasty distortion), and reduce the signal by 6 dB.
>
> 3) An output that basically drives pins 2 and 3 with opposite
>   signals (as above), but controls the difference voltage
>   between pins 2 and 3 rather than each of them separately.
>   This means for example that if either pin is shorted to
>   ground the difference signal remains the same.
>
> 4) An output that uses pin 3 as an _input_ with the same
>   impedance as pin 2. The signal on pin 3 is added to
>   the ouput appearing on pin 2. This provides ground
>   loop compensation even if the receiver is unbalanced
>   and connects pin 3 to ground.
>
> 5) Fully floating output - the signal is between pins 2
>   and 3 and isolated from the ground potential.

IMHO #5 is the most useful for computer-audio where there's a lot of
noise riding on the ground, and ground loop potentials.

Your #5 is best represented by
http://www.dself.dsl.pipex.com/ampins/balanced/balfig6.gif (
http://www.dself.dsl.pipex.com/ampins/balanced/balanced.htm )
" Simplified diagram of a quasi-floating balanced output, with its
essential trim control for output symmetry."

In terms of implementation, I  like using
http://www.rdlnet.com/product.php?page=54 (RDL STA-1 active
transformer pair) to implement  the "quasi-floating balanced out" from
a computer. The can float w/r/t their power supply. Perhaps the
ultimate "mod" is attach the output of your soundcard DAC to the input
of an STA-1 and get decent quality, floating balanced output. totally
bypassing the soundcard output stage and power supply.

Even better, find a $0.99 soundcard with an ICE1724 chip and a TOSLINK
port and take the $100-$200 you'd have spent on a quality soundcard
and get an external D/A converter, put the computer on a different
electrical circuit, and  interconnect with optical cable... best CMRR
around :-)

Niels
http://nielsmayer.com