What I DO notice is that the article is >10 years old and doesn't deal specifically with the proprietary asynchronous USB technology developed by Gordon Rankin of Wavelength - which appears to address the jitter and clock issues inherent in the technology. ....
correct, the article is old but I cited it just to show you the inherent flaws in the USB protocol & why Audiofreak32 stated that USB connections are "bad". You might have already know this about the USB standard so if it was a repeat, just ignore that info.
Simplistically, I believe that the TI TAS1020B part that Wavelength, Empirical & Ayre use has the ability to generate 2 clocks. Thus it allows itself to be the master & makes the PC the slave. Thus, jitter can be dramatically reduced using the clean clock on the TAS1020B IC. This was not possible with the other TI USB controller, the 2702. The TI 2702 IC used the data input to extract the clock. Also a bad side-effect of this was the USB controller had to wait for the PC to send the USB packets every 1ms +/- a certain amount to vaariable time. So, there was no exactness to when the data packets would start/stop. Jitter was very high. Thus, Empirical Audio & others (I think Wavelength Audio) simply gave up on this USB controller.
Both the TI 2702 & the TI TAS1020B use an on-board PLL and also FIFO buffers. So, I do not think that this was the technology improvement. I think that it was the dual clock generating ability in the TAS1020B that took the particular USB DACs you are referring to a new level of playback.
FWIW.