I need to work today, will be brief. Looked at the very first post then glanced at Alan's picture. Few comments -- please keep in mind this is after a very quick look as I am swamped, may be able to respond better later:
1. There is a length limitation for USB cables. I do not know it off-hand but you can do a search. Too long and signal integrity is compromised by excessive less (signal gets too small) and bandwidth roll-off creates high inter-symbol interference (ISI). Both lead to high jitter (random and deterministic) and can cause bit errors.
2. Within that limitation the manufacturer should not matter as long as the cable meets spec BUT there can be better quality connectors, cable (wire), shielding, etc. A $10 cable might be better in that sense than a $1 cable. I really doubt a $100 (or $1000) cable is going to make any difference.
3. I do not have a lot of experience with USB audio and do not use it in my system at this time. What I have seen in tests and in friend's systems are cases where the PC ground was very noise with lots of HF content. In some DACs (the box audiophiles call a DAC, not the actual DAC IC that I might design) the USB incoming ground is not well isolated from the analog ground on the other side of the DAC. Separating analog and digital grounds is pretty basic stuff for a data converter or mixed signal designer but may get overlooked by a PCB designer. The result is that "digital" ground noise from the USB link is injected onto the "analog" ground and you get a noisy output. I would not call this a ground loop, though there is obviously a current loop in there, but rather noise injection. Noise injection can happen into the ground (return) line, signal line, or both.
4. Noise injected equally into both sides (+ and -) of a differential signal is common-mode noise. The common-mode noise rejection ratio (CMRR) tells you how well such noise is rejected by the receiver (RF, opamp, transformer, whatever). Most opamps have very high CMRR at DC but it falls off rapidly with frequency. Circuits I play with must tolerate CM noise above 10 GHz, but an audio opamp probably has no rejection over 1 MHz or so. Noise at that point can get to the output. Alan's last picture shows a potential ground loop but the cause and effect in my mind is noise from the USB transmitter side injected into the receiver, and unless there is sufficient CMRR at that point the output can be noisy.
5. Isolation, e.g. galvanic isolation, and noise filtering (typically RLC networks, including the ferrites under discussion) can suppress the HF (RFI) the USB transmitter/PC injects into ground and common-mode into the signal pair. If that is the problem then a ferrite can help. Isolation can take the form of active circuits, passive components like transformers or independent coils, optocouplers, etc.
6. If a USB cable includes some sort of isolation and filtering then it could improve the sound if the DAC itself does not sufficiently reject the common-mode and ground noise. To my mind that is the fault of the DAC; other sites have measurements showing that some DACs lack immunity to ground noise. In my world such a design would never fly (literally).
7. A ground loop happens when signal (+) and return (-) currents take different paths. Ideally you want the signal and ground EM fields to be tightly coupled so they flow "together" down the cable. That way any noise happens to both equally and is rejected at the receiving end. If for some reason signal return (ground) is a higher impedance than some other path, like the chassis (safety) ground, then any noise coupled into ground can be amplified. For audio, this is typically 50/60 Hz noise from the AC supply, but of course other things can couple in. The long ground (return) path relative to the signal path means that coupled noise is no longer in phase with the signal and now that difference signal is amplified by the receiver. See e.g. https://www.whatsbestforum.com/threa...1/#post-375417
8. I think of a ground loop as happening when the signal (+) and signal return (-) (ground) currents take different paths. Noise injection happens when signal (+) and signal (-) (return, ground) take the same path but noise is injected into the signal (+), (-), or both. Obviously noise injection can occur in a signal path that has a ground loop.
What we are left with, or at least what I am left with, is two distinct noise sources. Noise (RFI - radio frequency interference) from the USB source coupled onto (injected into) the shield that creates common-mode noise at the receiver. If ground is not properly isolated and/or the receiver's common-mode rejection is not sufficient, then this injected noise can appear at the output. Noise injection does not need a long ground loop; it can occur right at the receiver (inside the box) or anywhere along the line. The second potential source is a ground loop but I tend to think of that as more of an analog signal problem rather than noise injection. Having the signal positive and negative currents take different paths, with ground typically being the longer path (loop), cause signal and ground to get out of phase and now coupled ground noise is amplified. See the linked thread.
When I responded I had the latter in mind and did not realize we were talking about USB cables. In a USB cable you can have either. So, a ferrite bead might help, if the problem is HF noise injection. It will not help if there is an analog ground loop, nor will it help (unless it is very big) if the problem is LF noise on the ground path getting into the analog (output) side of the DAC. Ferrite beads act like inductors, which have rising impedance as frequency goes up, so low at LF and high at HF (to a point). A small'ish ferrite will not generally do much for LF noise such as the 50/60 Hz noise coupled into what I consider and audio ground loop.
HTH - Don
Edit: I used a list but there was no space between items. Switched back to plain old text to add some space for legibility.