This sounds a lot like the problem I was having.
PC (mains earth) > USB (+5v, D+/D-, GND) > D10 > RCA unbalanced cable > XLR pin 2 signal, pin 1+3 tied to ground/shield > Active Monitor Speakers (mains earth).
Got a low level buzzing only really obvious when I turned the speakers up beyond normal listening levels or shoved my ear right next to the driver.
For me, using a twin core + shield cable terminating cable shield and ground together to RCA ground and signal to the RCA pin. At the other end the signal is connected to XLR pin 2 (+/hot) and the ground cable is connected to pin 3 only. There is no ground connected to the XLR ground pin 1.
Adding a 100 ohm resistor to the RCA ground end (as advised by another member here) would balance impedance with the output impedance of the D10 signal connection - but I didn't bother.
I am running a pair of 4 meter cables, with a bit of it coiled (naughty!) sat on top of the PC case and hanging right over a mains 4-gang extension lead, behind a router and modem... it's a filthy place but I get no noise at all. YMMV.
Other options are:
Optical output from PC, as Amir said. Optical has no metal cable connection and no ground connection. Proper isolation.
USB isolators (ADuM chip based) which may or may not have sufficient DC-DC transformers to provide enough power for the D50 from the original USB port - but it does have an external power input so likely not a problem. The downside, even if you got a well designed and well made one based on an ADuM chip, is that it will only do 24 bit 96 kHz at best. They simply cannot handle the speed of full USB 2.0 two-way traffic!
For example:
https://hifimediy.com/usb-isolator
Isolators based on a Silanna chip do handle high frequency audio and allow the proper full USB 2.0 data speed or more perhaps - but they ain't cheap. I've seen some use a proprietary FPGA setup.
iFi iDefender 3.0 :- A cleaver little circuit with an additional input for external +5v power. When an external PSU is added it intelligently cuts the ground connection between the device and PC. It operates at full USB 3.0 speeds, but is not an isolator, more a ground loop breaker.
Any of the methods involving shoving something in the USB connection would probably solve a ground loop issue but would likely cause some noise at 8 kHz. This is the noise from the actual data transmission timing or something or other. 'Spectrum spreading' or a term like that used in more complicated devices (like with the FPGA) could reduce this data noise.
Transformers in the audio path. I didn't like the idea of shoving something in the audio path. It's just a pair of coils and the alternating current in the audio signal (that's what it is) induces a current across a gap in another coil. Seems like a lot of potential losses in frequency response. Fine for in the car, but not at home. It's funny, but this is actually how the USB isolators work too - it's just that the coils are ridiculously thin and small and the data connection is two-way and timing is very critical to the transmission working or not.
Doing a dodgy test job or carving up a USB cable, breaking the shield and ground connections and seeing what happens. I had to connect the ground to begin with, or at least have a common ground between PC and DAC somewhere else the USB connection would not establish. Once established the ground can be severed, except for when sample rate is changed. However, DO NOT DO THIS for a long term solution beyond testing. If the Data lines fall too far beyond a certain voltage difference to the common USB ground, the USB chips could be damaged in the DAC or the PC motherboard.
Product designers could build in a solution, which would be AFTER the USB chip (XMOS) but before the DAC (ESS) chip. As the USB data is two way, it is a trouble isolating that. But the digital audio signal from the USB chip could be sent across a fast isolator. It could increase jitter (timing errors in the data) but would allow for full separating of USB ground from DAC ground. Singxer make something like this, with IIS outputs.
If you use a laptop (on battery or powered by double-insulated DC PSU (no mains)) then you won't have a ground loop and won't have an issue. The Raspberry Pi is fine in this regard. However, connect a TV set via HDMI and back comes some noise, actually a lot of nasty noise! The TV is mains earthed.
I suppose you could look into star grounding and that whole load of fun!
My little list of things to handle should I win the lottery and have a house built gets longer and longer...