There's been and continues to be much debate around source versus sink based decoding. I've seen very little input from what I would call people who are in a position of knowledge/experience to clarify. That said Charles Hansen founder/designer of Ayre Audio recently replied to the question -
I've heard some people claim that the undecoded streams are less affected by jitter than LPCM.
Anything is possible. But it's not clear to me why that would be the case. The problem was literally that the HDMI connector was designed before the rest of the system, and they didn't have a spare wire to use as an audio clock (although there is a spare wire that has never been used for anything -- go figure).
The video clock is sent, which for 720p and 1080i is 74.25 MHz and for 1080p is 148.50 MHz (or that times 1000/1001, thanks to NTSC). Then the source sends signals to the receiver telling it what ratios to use in the audio PLL to recover the audio clock from the video clock.
In the best of circumstances this is a relatively high-jitter way to generate an audio clock. So the receiver is stuck with a jittery audio clock. The question is "Will sending packetized bitstream information create lower jitter than sending PCM?"
I can't think of a single reason why this would be true. The bitstream information has to be decoded by the receiver. Then it is in PCM form, which the DACs chips can understand. So in either case the net result is to send PCM to the DAC chips. And the conversion to analog is controlled by the local master audio clock. And that is all that matters -- how low is the jitter at the DAC chip.
[b]So I would say that the claim of sending bitstream resulting in lower jitter is an urban myth.
Avii