I missed this thread when it was originally posted (2011), just saw it now. For higher-power applications of the TDA729x, the parallel schematic shown seems to be more robust than the Class-G schematic shown on the ST datasheet/appnote. Apparently, the Class-G schematic requires very precise rail sequencing, which is generally only possible by exactly following the ST schematic. Many people who tried to modify/improve that circuit ended up with smoke.
Also wondering if a parallel or bridged TDA2050 circuit is feasible - there's presently a closeout sale on the TDA2050 (which ST has stopped manufacturing), and prices will be very reasonable as long as stocks last.
On a related note, I'm revisiting the parallel MyRef-X2 variant with 2x LM3886 per monoblock which I prototyped a couple of years ago - the immediate motivation is that I tried a variant of one of the alternate Rev C compensations today, with some upgrades to various passives, on one of my earliest Rev C builds, and the audible sonics turned out to be very well-balanced. That leads me to believe that this will work well with similar improvements to the MyRef-X2 as well.
