There are a few points worth making here.
The problems with S/PDIF are intrinsic to its design. Even with as close to a perfect a specified S/PDIF feed as one could imagine, one with no jitter, wonderfully perfect waveforms etc etc, the actual nature of the beast leaves you with intrinsic problems. The most fundamental is the difficulty in recovering the clock without it being contaminated by signal correlated jitter.
Now there is a lot of experience with this, and to a pretty reasonable extent it is solvable. But not without serious effort.
USB audio is an almost utterly different beast. But it too has intrinsic issues. The worst of which is the lack of any reference clock. So we must recover a clock from the received data here too. This leads to problems, but for different reasons to S/PDIF.
I2S is a better encoding of digital audio, in that is separates the clock from the data. So once we have a clean clock - and getting that clean clock may be quite a trial - we should be able to avoid it becoming contaminated with jitter - especially signal correlated jitter. But the I2S output from any receiver, be it an S/PDIF in a conventional system, or a USB device such as a PCM2706, will only be as clean as the internal capabilities of the receiver. If it were S/PDIF the likelihood of signal correlated jitter on the I2S clock will be very high, with a USB receiver, more likely just lots of auto-correlated noise in the jitter spectrum left over from the vagaries of the USB transmission.
If you have not read the lovely story of the design of the PCM270x series, you need to. Understanding why the SPACT design does what it does is important to understanding the tradeoffs discussed.http://www.planetanalog.com/showArti...cleID=12801995