previous post was probably a bit too terse...
The problem here stems from mis-use of terminology, not unlike the ever popular phase vs polarity. If you know what you are doing, then it really does not matter what the label says, but these tiny errors cause problems when trying to explain things.
An input can never be balanced! It can be differential, which means that it takes the difference between the two input pins as the signal, no external reference is required. In doing so, it will add signals of opposite polarity (the signal we care about) and subtract signals that are of the same polarity (the noise.) It doesn't matter if there is a signal of interest on both pins, if there is not then it simply subtracts nothing from the signal that is present on one pin.
An output, or more correctly a source, can be balanced, or it can be single ended (un-balanced). If there are two signal pins then it is helpful if the signal of interest appears on both pins, but the polarity on one pin must be opposite the other. The impedance from each pin to ground must be equal, or really close to equal for the differential input to operate optimally.
If you look up the IEC definition for testing CMRR you'll see that the test measures the noise with the two legs optimized, and then with each leg "out of balance" by different resistances. In the case of a well designed transformer the CMRR suffers, but not by a lot. A standard three-opamp input stage suffers terribly. The InGenius chip fairs pretty well... much better than I ever expected!
So if you have a single-ended output you can wire it as suggested in Rane Note 110 (everyone should have a copy!) and you will still achieve some benefit from the differential input stage - if it is designed well.
Why? Well this is where the wire comes into play...
So we have two connections at the source, one goes straight to ground, the other goes through some small impedance to ground (these days the source impedance of most outputs in 50 ohms or less, and 10 ohms is not uncommon for brave/foolish designers<G>!) So that actual imbalance between the two pins is already very small.
These pins are then connected to the two input pins, either through a twisted pair or through a coaxial cable.
As mentioned earlier, shields don't do much to protect a conductor from magnetic fields. So in effect, both the signal conductor and the shield are exposed to the same magnetic field. Distance and orientation of the source of the magnetic field play a role, but for this discussion we can ignore them.
The difference in impedance to ground does play a role here. The magnetic field impressed upon the conductor and shield is the same, but the resulting current flow (and therefore voltage) will be different - see Ohm's Law!
The next step up is to connect the pins to a twisted pair (shielded or not). Shielded pairs work well because the twist causes the two conductors to be exposed to almost exactly the same amount of magnetic field. Pretty cool eh?
If one of those conductors is connected to ground at the source then we have essentially the same conditions as if we used a coaxial cable. The field strength is the same, but the resulting voltage differs because of the imbalance between the impedance to ground for each pin.
The best solution is to use two pins that have the same impedance to ground - AND signals of equal magnitude and opposite polarity. Note that the later condition, properly referred to as signal symmetry, is not a requirement.
Signal symmetry provides a 6 dB advantage (all other things being equal) in S/N ratio because we end up with twice the voltage at the output of the differential input. That's it, that's what happens with a symmetrical balanced source feeding a differential input via a pair of wires.
In terms of immunity to impedance imbalance (or mis-match) transformers are best, followed surprisingly closely by the InGenius chip, and pretty much every other differential topology falls far below these. The worst is a single op-amp, because it presents widely differing input impedances, no matter how clever you are. Somewhat better is the classic instrument amplifier, where each pin connects to the inverting pin on an opamp (some designers will use the non-inverting pin because the lower effective impedance provides some noise immunity) and then these two opamps drive a difference amplifier stage.
For the "why" transformers work best I recommend reading
Bill Whitlock's chapter on transformers in the AudioCyclopedia. He has also published a paper on
how the InGenius chip works. Actually, I'd recommend reading pretty much everything at the
Jensen web site!
Of course it isn't quite this simple, there are problems with Shield Current Induced Noise (SCIN) - indentified and described by Jim Brown (
here, and
here), and the infamous Pin-1 Problem, indentified by Neil Muncy and written about extensively by both Bill Whitlock (see the Jensen site) and Jim Brown(
here and
here)
Sadly Neil's paper is only available through the
AES. Other AES resources include
Standard AES48 and John Windt's paper on the
Hummer Test.
All of these topics, and of course the whole power and grounding issue, are tightly interrelated, so it is tough to talk about one without the others.
All this to say that yes, connecting a single-ended source to a differential input with a shielded twisted pair as described in Rane Note 110 will provide some benefit... but it is not anywhere near optimal. So the question remains, do you need optimal? And that's something I can not answer<G>!