I'll get the multimeter out in a while.
I already tried the cable above splitting D+/D- and +5v/GND to two different locations, and it did not work for me.
The D10 turned on, but showed the dashes on display, eventually going into standby mode (single dot?), there was no recognition in the PC.
What about making that same cable, but shorting +5v and GND on the PC (data source) with a resistor? Would this 'trick'the PC into thinking a device is present, or would it just result in potential damage?
Ah, now this looks interesting.
https://www.bamlog.com/usbisolator.htm
In particular, the author states that the device may still need to be left un-isolated on connection, so that it can enumerate/initialise. Once that is done, the toggle switch can then isolate the power to use external.
I have demonstrated that the red +5v line to the Host device is not required when the DAC is powered with external PSU. So, really, all I would need to do is to just break the USB ground to Host _after_ the DAC is initialised, basically when I see a sample rate on the D10, then the ground loop (and buzzings) should just go away.
Not sure what values to use, and whether or not the cap. is in the right place, or required at all.
Another version, same thing more or less.
http://www.madscientist-audio.com/gbu.html
However, is that a good idea? USB is designed with different lengths on the tracks of the connector. The Data lines are the last things to connect, after ground and then power.
I am now heading more towards the original topic title, even if I am a little skewed (not clean power, rather clean ground/brake ground loop).
http://zone.ni.com/reference/en-XX/help/370592AB-01/digitizers/513x_ground_loop_noise/
...and for those who missed it:
Silanna ICE08USB is the USB2.0 (up to 480mpbs) alternative to the ADuM4160 USB1.1 (up to 12mbps) galvanic icolator.
It's seen in the UpTone ISO Regen (which did nothing in Amir's test - but he didn't have a ground loop) and also CoolGear's Isolator.