I know Ted has discussed this a bit before and in essence his opinion is that all of the inputs can sound very good, however designing them to sound good is easier or harder to do depending on which input we’re talking about. He mentioned that I2S is one of the easiest ones to get to sound good. Here is a post where he discusses this. Personally I’d be using toslink if it had the same sample rate/bit depth (and DSD) input options as I2S and USB had. It doesn’t, so I use I2S as my choice of input. I think it sounds better than all of the other input options, though that may reflect on my source component.