Well I finally was able to implement one ~ “final” (?) electrical mod to my DS, and it might actually be the biggest of all in fact based on what I’m hearing. May be we saved the best for last. So felt I must share here.
So meanwhile by now we’ve done the more obvious things. In my case a progression from stock output transformer (OPT) → XS4400 OPT → no OPT w/ some super high quality caps → mu metal OPT → amorphous OPT. And soon now we will have a pair of amorphous w/ pure Cu windings. I think it has earned it 
Suspect that, especially considering the footage involved we’ll benefit audibly from the “quietest” conductor we can get in there. Though if I had the choice it might not be the Cardas. But it’s all we can do for now.
And by now we’ve also substantially beefed up (in the high end analog circuit sense) each of the (8) local op amp PSU’s along with the several upstream nodes that feed them. We’ve also beefed up the DSD bit slinger PSU as well; personally I consider that part of the analog in this DSD case .
But to the mod here- It’s been in the back of my mind for awhile: what exactly should we do about the Vocm node in the op amp circuits. Thinking out loud here, that node defines the exact center reference point of the differential output waveform. Thus it’s analogous to the VSS node in single ended circuits. Seems to me it must be absolutely rock steady, certainly through the audio band at least. Any wiggle there is unwanted noise then and passes straight through to the output 1:1 with no attenuation. I think we need that dead quiet.
The AD8132 datasheet has a small amount of info on Vocm, and seems there are a few statements that support the above thinking (my emphasis):
In cases where more accurate control of the output common-mode
level is required, it is a best practice that an external source or
resistor divider (with RSOURCE < 10 kΩ) be used. The output
common-mode offset values in the Specifications section assume
the VOCM input is driven by a low impedance voltage source.
BTW the DS design follows the datasheet basically and has a precision resistor 50% voltage divider that creates the Vocm node with some ceramic bypassing in (what I assume is) the ~ high speed sense. A resistor voltage divider generally creates a ~ high impedance output node; only buffering or bypassing brings it down to low impedance. The datasheet seems to imply AD used a buffered, low impedance source to drive Vocm for their best specs…
As an editorial aside here, my opinion, I do not like ceramic caps anywhere near precision analog audio circuits 
Generally speaking they have audible issues- voltage coefficient and piezo effect, basically. In this case I probably would agree though, likely not a big factor (except maybe the Vocm node (?)). But I’m not going to bother trying to rip them out here
Anyway, key point- I think that Vocm node likely needs to be as immovable (ie low impedance) as possible through the audio band; so it really needs an analog class bypass.
So yesterday I added what I think it needed, and the results seem amazing to me so far. Not knowing what to expect at first, after a few Tidal songs the big word that literally leaps into my mind is… TEXTURE… Wow… so much texture and preservation of low level detail. There has not been this level of texture in my system before. It seemed almost overwhelming at first; my brain had to adjust. The sound is correct, natural. That’s what I’m after lately. And as we push forward there can be no going back
The Concept 500 also help in this as they specialize in transparency, correctness and delivering, revealing what they are fed as well.
To make sure I’m not just imagining things, I’m really hoping another person might also try this and report back.
Below is an image showing the Vocm dividers. Please note the Vocm node is not between the resistors as you might assume.
What I did was, I already have a PPS SMD bypass directly across each resistor R_A and R_B. But the real mod is the analog bypass- I placed a Bi-Polar (BP) 47uF Muse across R_B. This bypasses the Vocm node to local VSS. I had these on hand with 5mm lead space which makes them ~ suitable here. And Yes, BP is better sounding than polarized in general. If I had same in 100uF I would have used that instead of 47uF.
The best sounding electrolytic I know of is Kaisai. And I have extensive experience using those. But I think we can get close to that (enough for now) with the BP Muse at much more reasonable cost.
And since this cap affects the signal directly (basically), use only the best. Don’t bother with anything less. If you skimp out and use anything less, we cannot really compare.
And if you have the patience, likely even better would be a 47uF cap across each R, A and B.
Place the “negative” cap lead on the lowest impedance side of the resistor. In the case of R_B that would be the VSS node on the LHS of it in the image.
Looking forward to hearing what others may experience.
Thanks, T
I’ve added some links to quick cell phone clips. Don’t expect the world but maybe some can get a sense of what we’re hearing here. I’m not used to using Google drive but maybe this will work: