Indeed.Your scheme can't handle common-mode gracefully unless its very small.
I see no reason for an extra Diff to SE converter and back to Diff.
The error in the shown circuit diagram is that there are no feedback resistors and that the two resistors connected to the minus input are grounded in the middle.
With the corrected diagram below, Common mode rejection is excellent, and since the speaker is floating, no sound will be produced when offering a common mode signal.
I regret that the first scheme was drawn wrong. This last scheme is closest to the real one. Practically speaking, there are two ordinary single ended power amplifiers on each channel.
I disagree, there is no need for this additional complication.I suppose my question would be whether you want to do common mode rejection on the speaker outputs. Seems to me that a good sized spike of noise could easily clip both amplifiers causing all sorts of havoc. It actually makes sense to invert and add the XLR signals in the conventional manner then invert again to drive BTL loads, just to avoid this.
The conversions are not all that difficult...
If you look at the thumbnail below you will see the advantage ... the inputs on the top traces are showing a lot of common mode noise but once through the summing resistors it's completely gone from the middle panel. The bottom panel shows a nice clean differential output that could be fed to your power amps.
It's ok that ?
I disagree, there is no need for this additional complication.
The circuit diagram as shown has a perfect CMRR, see image below.
V(out+) and V(Out-) are common mode polluted to the main amp's input.
At the speaker output, in this image divided by 40 to get the same scale, this common mode signal is completely removed.