Because of USB’s ability to slave the server’s clock to the Dac’s it should have inherent advantages over other connections, particular where the DAC has superior clocking. Alas actual experience often differs. USB seems prone to ground level and RFI/EMI level incursions and the actual USB control often needs reclocking as well.
I have been through a long journey improving my USB link. After trying and rejecting conversion to SPDif and AES/EBU I started by inserting an Intona Isolator plus Intona Reference cables. Not entirely happy I added a Singxer UIP1 Pro decrapifier and reclocker, The resultant improvements were substantial in terms of reducing background noise, clarity and fluidity.
Out of sheer curiosity I now added an Ifi Purifier3 between server and Singxer to the chain, i.e. server-> Ifi Purifier3->Singxer UIP1 Pro supplied by Sean Jacobs LPS->Intona Isolator->DAC. Lots of boxes there, yet: to my substantial surprise there was another significnt lifting of veils. The conclusion seems that multiple moats and filters on USB are additive rather than redundant. The subject of noise incursion into digital links in my mind needs substantial further investigation; its negative effects are substantial on D/A conversion