I have followed the recent series of posts with some interest. Jitter is a vexed issue, not in the least because measuring and characterising it is very hard. Finding its sources is even harder.
I should clear up one misconception: the main source of transport/cable related jitter is that the clock signal is buried in the S/PDIF data stream. This is bad. Because the clock is extracted from data that is encoded usign a bi-phase mark system, skewing (due to cable properties), fluctuations (caused by any number of things, including power supply loading, RF, crosstalk etc) will distort the timing of the edges. Recovering the o's and 1's is trivial compared to recovering them at the right time. CDs use eight-to-fourteen modulation, cross-interleaving and some parity bits, so a clean CD will give fair robustness to errors, but jitter remains as the primary goal of such approaches is signal integrity not signal timing. So everything you guys have said is true, however, jitter is yet another pile of sludge on top of this. Let me charge on here...
The main jitter problem comes from transport and output premamp power supplied modulating the on-board clock and other digital circuitry causing jitter. The classic solution is to put the DAC in a separate box with its own supply. However, the DAC receivers use PLLs to recover and smooth out the clock, which have to allow for a degree of slop in precise transport speed (the standard is +/- 400ppm!) and jitter on the incoming clock - thus the transport does affect the recovered clock, albeit somewhat diminished dependeing on how good the PLL is. The PLL's performance is a basic limiting factor in DAC jitter. All else being equal optimising the PLL gives you a major performance improvement. The remaining problem is that signal dependent artifacts can still intrude on the recovered clock in a number of ways, and since the two systems are effectively synchonised these artifacts are of the worst kind - signal correlated. This tie can be broken by "desynchronising" the systems (transport and DAC) which can be accomplished using a FIFO and a few other tricks. The rest is just good design: power supply regulation, layout, optimsed component choices and careful attention to grounding and RF.
It is true that the transport and cable will always have some impact on the final result (usiggn the current interface that is) but it is possible to reduce this by a large factor if not an order of magnitude or two, and that's all that's needed to put the effects below the noise floor.
As far as errors go, you can use somethign like exact audio copy to make a crude log of how many and what kinds of errors are on discs. The worst error you get on CD is a "fatal" one requiring interpolation. Philips used to make a test disc which had tracks with deliberate errors of different kinds so you could check the transport and error-correction system. Larry is spot on. There are a lot of dirty transports out there, but it does take a lot of data corruption to actually corrupt a sample. I will endeavor to do some tests to find out how much corruption is required to actually mess up a sample. Could take me a while.
Hope this helps. Pardon the ranting.
T.
