My ANK DAC is a noisy bastard - not sonically, but RFI/EMI, so much so that I lashed-up some grounded steel+neoprene+brass isolation platforms to reduce the pickup by kit above and below it in the rack. Before that, you could hear music playing from the pre when the power amps weren’t switched on! The platforms helped enormously.
But that’s not the story. Like a dog returning to its vomit, this compelled me to go back and look again at mains filtering for the DAC. What’s spewing onto the airwaves is also getting piped along connected wiring. Some experimentation with additional off-the-shelf modular filters only made the problems (and subjective SQ) worse…
I tried the thing with no filtering at-all, just my screened mains cables, and that changed the presentation - both for better and for worse…
Most commercial passive mains filters mix inductors with capacitative shunts across L-N and across L-E and N-E. The first should cancel common-mode noise, the second offers a path to Earth for HF interference. Both have their issues, but the latter especially can simply make it easier for strong noise sigals to proliferate throughout a system. It’s an engineering solution that looks at a problem in isolation, and ignores wider issues - if you’d got one noisy machine in a factory fifty years ago that was upsetting your brand-new fax machine: sorted. Now, not so much.
What industrial filters don’t use much of is inductance, at best you get a very low L common-mode transformer, because copper wire is expensive plus they take-up a fair bit of space. There are also potential issues with current-generated heat as inductance values rise, especially with low-CSA wire windings. You also pretty much never see any differential chokes included.
With such a noisy bit of kit as the ANK DAC, I decided to try making a standalone filter that would combine a DC blocker with multi-stage inductive filtering - and so ordered a bunch of cheap 50uH common-mode chokes from thiefBay.
Tie two legs of these together and you have a differential mode choke - keeps the part count low! The result of hastily lashing these onto a bit of perfboard is this thing of Great Beauty 
I’d agree that the enclosure may not be fully production-ready…
The caps and diodes in the middle are just to block DC offset. LNE all have a differential choke on input; LN then have one common-mode before the DC block and two after. E has a second differential choke just to get numbers up. This config also balances impedence at 0.15R at 120Hz for each leg, something which climbs smoothly with frequency, as you’d hope.
What matters (and why I’m bothering to post to universal indifference), is that it works absolutely brilliantly at a subjective level - the system sounds better in all ways, resolution is improved, but most importantly treble brightness and glare is significantly reduced.
I didn’t really expect this, I’ve got a selection of pretty reliable and effective mains filters already, but for this pernickety DAC, this is just what was needed, and the result is a fair bit better than anticipated.
Just the small matter of it being deadly to take care of at some point…
