Differential Balanced Sound Quality

Balanced connections are all about eliminating external noise that gets picked up in cables etc. They also pretty much deal with ground loops as well. So if either of those things are a problem in your system then using balanced connections is effective way of dealing with them. You also get a 'free' 6dB of gain when the balanced signals are mixed back to one signal... but I don't think gain is in short supply with modern equipment.
The bit that a lot of people are unaware of is that balanced inputs are actually noisier than single ended (RCAs etc.) in that they utilise relatively high value resistors that introduce their own (johnson) noise into the signal.
So if you don't have problems with interference or ground loops then single ended will be quieter. It's really down to your own system as to which is better.

When I read your first sentence I thought you might have been talking about bridged amplifiers vs single ended... that's another subject if you did.

The bit that a lot of people are unaware of is that balanced inputs are actually noisier than single ended (RCAs etc.) in that they utilise relatively high value resistors that introduce their own (johnson) noise into the signal.

This statement is problematic. There’s no reason why any such ’high value resistors’ be used in a balanced circuit that aren’t also in a single-ended circuit.

There is a reason: In most consumer (and professional) amplifiers the balanced signal is dealt with at the input by a differential amplifier necessitating the need for a series resistor on the inverting input. In the worst case the diff amp is the load seen by the source so the resitor needs to be large enough to present a reasonable input impedance (that's the 'relatively high' part).

A better solution is to use an instrumentation amplifier which buffers the input so all you need to worry about is the ability of the buffer to supply current to the diff amp in which case the series resistor can be smaller. But in both cases there needs to be a series resistor that is larger than that required for a single ended input fed into a FET for example. 
A transformer input is obviously different and more of a rarity but very beneficial for noise reduction and does a better job at dealing with RF/EMC.  However transformers are far from perfect, don't have great LF linearity and HF response can be iffy, so no panacea.

Another option and perhaps the one that the OP was alluding to is feeding the balanced signal right through to the speakers using a bridged amplifier. I am less aware of the pro's and con's of doing it this way but I'd be interested to know what the measured CMR is of such systems as the gains of the two halves of the bridge would need to be very well matched.

My point was that balanced isn't 'better' by nature, just different and if you have a noise free environment with no ground loops it could be slightly worse.

In most cases my advice would be to go for balanced as there are likely to be more people who would benefit from the removal of noise at the input than would notice the additional noise from the circuit.

I apologize if you see this as being patronized, that was not my intent!

Cool... no problem.

Take a look at the first hit (this is an images search). The first image shows both what you’ve described **and** what I described in the same image, the latter of which does not have the Johnson noise issue.

Absolutely, agreed... I was talking about the topology shown in the first image.

IME the statement is only true if opamps are used *and* the internal circuitry is single-ended

Yep... op amps are the usual way of doing it and that's what I was talking about.

BTW love the look of your amps and I can see the passion you put into them.

There's no such thing as 'cold signal'.

The term is commonly used both in text books and the professional recording environment as shorthand for the inverted signal in a balanced pair... It doesn't confuse me and I don't need to be patronised.

If you do a quick google search on differential amplifier you will see that by far the most common use for this term is exactly what I have described and this is what is used in the majority of equipment with balanced inputs. 

The conversation flow of this thread is about the pro's and con's of balanced connections and what I have said is correct.

@atmasphere Happy to try to explain, it’s a shame that there’s no way of posting a picture...

So if I start with a conventional amplifier block with an inverting and non-inverting input and a single output. A single ended amplifier input might be a 100Ω series resistor followed by an RF filter & DC blocking capacitor, in non-inverting mode the input impedance is set by the resistors to ground at the input so it’s not difficult to maintain a high input impedance alongside a low thermal noise from the series resistor. The actual impedance of the non-inverting input is so large that it can be pretty much ignored.

If you take the above example and feed the cold signal into the inverting input, the series resistor on the cold input will dictate the maximum input impedance as the current will be flowing into the virtual ground at the summing point. So 100Ω is now out of the question. You might for example choose to go with 10kΩ series resistors on both inputs, that’s 20dB more thermal noise than 100Ω.

When I look at your amplifier I see that you have two outputs and I suspect that is the source of confusion... at what point does the cold signal get inverted?.. or does it connect to the negative speaker terminal?

Edit: I just did a quick google search to find a picture... I know nothing about the site and I've not read the content but the schematic in the header is what I'm talking about. The cold input current flows into the summing point (where the Va label is), so R1 sets the cold input impedance... in fact the impedance will be lower than R1s value but that's beside the point. If we change this to single ended with a gain of 1/1, R1 becomes open circuit and R3 is a dead short. The input impedance is R2 + R4, which means R2 can be low and R4 can be higher and the thermal noise is calculated from the voltage divider.