Transmission standards without an embedded clock, like NRZ
(»SDIF-2«), may face a serious problem: word-clock
(i.e. the frame) and audio data (i.e. the contents)
may deviate from each other, owing to different
propagation delay times when travelling along different paths.
The SDIF-2 standard is extremely touchy in this respect, as not only
the frame, but also the bit clock is derived from a word
clock that may be separately guided. With a mismatch as low as ±1.6%,
the bit assignment is lost. Result: the best-case is a signal louder or
lower by 6dB, in the worst- case, however, the signal may crackle and clatter
heavily, when the transitions between two bit-valencies are scanned,
instead of the steady states midway between adjacent transitions.
Compared to SDIF-2, self-clocking transmission codes like the
biphase-mark (»AES/EBU«) are well-behaved. Problems
occur only with deviations greater than ±25%, and even then only
when, real bad luck, a sample transition gets damaged. A minute
step in time later, the next word will be captured instead, which will
change nothing but add 20µs of transport delay.