FYI, USB audio does NOT feature error correction. It is not like your printer that will re-read the corrupted data (USB bulk mode).
Does it matter? Probably not, unless there's a heavy activity on the computer, there should never be any data dropout (I read a paper that said it would be in the order of one audio sample every few hours / days).
If you think analog cable make a different, consider how much more bandwidth a digital signal requires. Theoretically, it is infinite (square signal). We're dealing with transmission lines at this point: much more critical than a 5-30kHz audio signal.
Yep, fair enough, it is one of the modifications required to make USB stream audio better e.g like UDP and not TCP/IP. This is why the playback software, FIFO and buffering strategy matters. Ideally you'd like to buffer the entire song in memory and optomise the flow to the audio receiver without interruption, where you would have a small memory buffer that is will be continuously filled with blocks. The buffer will be emptied using the on board receiver side clock which should reduce the impact of jitter from the transmission side. Yes, a square digital signal requires more bandwidth (a square wave contains lots of unwanted EMI, therefore most signals have a trapezoidal like look), but USB uses differential signalling which as long as your cable meets the USB 2.0 specification should not matter.
Edited by daerron - 6/4/14 at 8:43am