How much advantage did you find by adding shielding of the XLR over just using a differential input w.r.t. noise?Almost none. I ran unshielded tonearm cable in my home system for some years. You could crank up the volume and no hum or buzz from the cable, even if grasping it or moving it around. You do still have to ground the tone arm and it works best if that wiring travels with the signal wires. One advantage here is this allows for much lower capacitance in the cable.
As roberttdid said about differential filter "it seem obvious", but it escaped me completely, being stuck on "matching".@kijanki I think a lot of people think of matching 'equal but opposite' circuits when they hear the word 'balanced'. EQ circuits might be the best example of that. I once went to see a demo of the new (at the time) Mark Levinson preamp (IIRC the ML-29) which was their first balanced line preamp. But the phono was single-ended. When I asked why, I was told that matching the EQ was the problem the 'engineers' were worried about weird effects if the parts weren't matched really carefully! That was when I knew they hadn't tried it at all.
At any rate if you want to do balanced right with active balanced circuitry, the way to do it involves differential circuitry. That's how you develop power supply noise immunity (Cross-Mode Rejection) and Common Mode Rejection. (As a side note, the CCS is critical and frankly, most CCS circuits I see are pretty terrible.) But what is less obvious but also important is the simple fact that you should do wiring in differential mode too- common grounds and common power supply points being the most obvious examples, and of course the EQ can be done differentially as well. Now we use passive EQ for our phono (based on the formula of Stanley Lipschitz) so imagine dual EQ networks, one for each phase; this made it go easy.
Those networks should use a common ground of course, but if you think about it, you don't need the ground at all. And if you got that far, then you can see that resistor values are in series as are capacitors- meaning that only one resistor need be used, only one cap (of half the value; you can see where this is going) and now there's one network instead of two. Much, much easier and all you're doing at that point is trimming to the values you need, rather than a Sisyphean task of matching to some extreme; pointless when you have tubes or semiconductors that won't come anywhere close to a similar match. A nice result is you don't have to do crazy amounts of tube matching yet the EQ will be spot on from both phases even as the tubes age. As far as I know, we were the first to do this (1990) and I think it was only about 8 years ago before I spotting any circuits that did the same thing- apparently its not common knowledge even today.