Originally Posted by Mycroft
I can't say I understand this. Couldn't the receiver just have a buffer, thereby making the clocks independent? Or does that cause syncing problems?
You want the easy answer or the esoteric one?
Here is the easy answer. Yes, you can use a buffer and then run an independent clock on the output of the buffer to drive the DAC. But now you buy yourself a new problem. The input clock and output clock are independent of each other and can drift in different direction. Should the output run faster than the input clock, you eventually run out of samples to play! So you wind up interpolating data to stall until you get more input and this can impact audio quality. If the output runs slower than input, then you start accumulating samples as you run behind more and more (think what would happen in a 2 hour movie, even if your clock is very close to input). This means that you can run out of memory if the buffer is too small, forcing you again to do some kind of sample rate conversion which again impacts audio quality. Now, you may be tempted to put a huge buffer in there given cheap memory these days. But then you wind up accumulating latency. If the buffer becomes a few seconds, you will have problems with UI responsiveness to things like pausing audio.
There are adaptive de-jitter circuits which try to deal with above but still, the issue is hard to deal with in absolutes.
Now the esoteric answer. The same LIM (Logic Induced Modulation) distortion which got us started in this topic, applies here the same. The input jitter can cause variations in the power supply circuit which could be transmitted/reflected in the output clock, causing jitter there! Yes, I know, hard to imagine but if you want to have 24-bit noise floor, it is remarkable how little jitter budget one can have to achieve such accuracy.
All of this occurs because we have a real-time system. In data communication circles, we don't have these problems because the final resting place of the bits is on some storage device. As soon as you require that the output be synchronous, then the problem gets a lot more complicated (and less understood at times as we don't really know all the audible effects of jitter).