That is not at all what I am implying… When a digital audio stream is altered it can happen a few ways:
- bits are changed and/or lost and caught by error correction techniques causing a retransmit
- bits are changed and/or lost and not caught by error correction techniques so they are allowed to be played
In the case of #1, if the errors are caught, a retransmit can be requested and if the retransmitted data arrives fast enough that the buffer is not emptied, then the resultant stream is still perfect with no error.
Checksums and sequences numbers can prevent #2 easily, but #1’s retransmits can still take too long to arrive. This can result in a buffer emptying. If the buffer is emptied, you will hear a loop of the buffer or zeros or something else bogus. The sound wave has been damaged; the DAC will not find a continuous audio wave, and will output very unexpected results.
This usually can be heard as a large click or pop, or as silence. A non-networked example of this that we have all heard is a CD that skips. That just means it couldn’t read the data off the CD (and it can verify that the data is valid using the same techniques listed above) before the buffer ran out. There is no “quality loss” when a CD skips… it’s just an “obvious error”. It’s not like the sound got muddy or lost fidelity in some way, it just went to shit.
The worst of the worst ethernet cable would result in the bits being damaged/lost – a good protocol can catch #2, so #1 is the case to worry about. That case would result in retransmits, which if the cable was bad enough, wouldnt arrive in time in a verifiable manner, meaning you would hear “obvious errors”, and not fidelity loss.
The reality of these retransmits is that they happen fast and buffers are relatively long, so even if your network is shit, things probably just work fine. The digital stream can not be altered along the way. That’s the point of making it digital.
Note that ALL of the above is purely in the digital part of this signal path, and claims about a bad cable, noise, ground loops can not affect it, because digital is built on mathematics, and not the realities of electricity. Either it arrives there good and verifiable, or it does not. This binary good or bad nature of a “reliable digital stream” is what drives the “bits are bits” guys nuts when audio guys talk about digital streams being affected by anything.
The claims about noise, ground loops, or whatever else is purely in how that digital stream is interpreted into analog, which is not a digital process. That stuff can not be verified – thus all the trouble. This is what drives the audio guys nuts when the “bits are bits” guys tell them they are old and not versed in information theory.