The isochronous data transfer is basically a one-way packet sent from PC to DAC via USB. It is sent 1000 times per second, and does not have error correction. This is the basic USB audio that most of us use. If a packet has an error or gets dropped, then there is a glitch or pause in the output. No buffering of data. Think of it as sort of a pulsed S/PDIF stream. Any glitch is immediately heard in the output, no error corrections.
The output word clock from most of the USB chips on the market is derived from this 1000Hz packet rate. Some xtal in the PC sets this timing reference (can anyone say jitter?). The USB receiver then uses an asynchronous clock to catch the data. It is then re-clocked (from packets to continuous I2S or S/PDIF) using a new clock derived from a PLL based synthesizer that uses the 1ms packet rate as a reference. Both the USB clock and 1ms clocks impact the jitter performance. The onboard PLL does help to remove some of the reference jitter. Results can actually be pretty good.
The other USB mode commonly in use by pro-audio companies (Edirol, M-audio, etc.) is to create their own custom USB interface. This require special software drivers but allows them to run other data rates (24/96), multiple channels (more than stereo), error correction, or even clocking schemes where the receiver is master. It allows for a lot of potential performance.
Most of us use the PCM270x or PCM290x series of chips from TI. They are easy to use, and relatively plug-and-play with both XP and OS-X. Our products are limited to 16 bit 44.1k/48k resolution. At the moment. Someday I expect TI and the operating system folks to take this to the next level (I could be wrong).
That said, in my opinion, the isochronous USB audio isn't really inherently superior to S/PDIF. They both have issues. The both can have jitter. To me, neither is vastly superior over the other. In fact, they both can be implemented quite well. My solution to both is to perform the reclocking afterwards. That is, right at the converter output. This essentially compensates for most of the signal deterioration incurred in the data transmission between boxes.
To me, I find such jitter-reducing reclocking to make a big difference in the sonics. Everything becomes smoother and more relaxed. In a way, the music almost slows down. Any harshness or edge or agressiveness disappears. I guess the most noticable attribute is that long-term listening fatigue is eliminated. For me anyway, that's the way it is.
jh
The output word clock from most of the USB chips on the market is derived from this 1000Hz packet rate. Some xtal in the PC sets this timing reference (can anyone say jitter?). The USB receiver then uses an asynchronous clock to catch the data. It is then re-clocked (from packets to continuous I2S or S/PDIF) using a new clock derived from a PLL based synthesizer that uses the 1ms packet rate as a reference. Both the USB clock and 1ms clocks impact the jitter performance. The onboard PLL does help to remove some of the reference jitter. Results can actually be pretty good.
The other USB mode commonly in use by pro-audio companies (Edirol, M-audio, etc.) is to create their own custom USB interface. This require special software drivers but allows them to run other data rates (24/96), multiple channels (more than stereo), error correction, or even clocking schemes where the receiver is master. It allows for a lot of potential performance.
Most of us use the PCM270x or PCM290x series of chips from TI. They are easy to use, and relatively plug-and-play with both XP and OS-X. Our products are limited to 16 bit 44.1k/48k resolution. At the moment. Someday I expect TI and the operating system folks to take this to the next level (I could be wrong).
That said, in my opinion, the isochronous USB audio isn't really inherently superior to S/PDIF. They both have issues. The both can have jitter. To me, neither is vastly superior over the other. In fact, they both can be implemented quite well. My solution to both is to perform the reclocking afterwards. That is, right at the converter output. This essentially compensates for most of the signal deterioration incurred in the data transmission between boxes.
To me, I find such jitter-reducing reclocking to make a big difference in the sonics. Everything becomes smoother and more relaxed. In a way, the music almost slows down. Any harshness or edge or agressiveness disappears. I guess the most noticable attribute is that long-term listening fatigue is eliminated. For me anyway, that's the way it is.
jh