[
The S/PDIF data rate is 12.288 MBits/s (32 bits X 2 channels * 192000 samples/s). I'm not sure whether a channel with 12.288 MHz bandwidth is able to transfer such a data stream - the rectangular pulses would be reduced to sine waves and the eye pattern may look bad.
When I designed my DIY digital patchbay I tested the coax IOs with 44/16 (data rate = 2.816 MBit/s) and AFAIR the bandwidth was around 6 MHz. So a factor of 2 might be required for a safe transmission which means about 25 MHz for 192 kHz sample rate.
And coming back to the OP: For digital audio transmission one should always use proper cables with the specified impedance as required by the standard (75 Ohm for S/PDIF, 110 Ohm for AES/EBU). There is no reason to use other cables since proper cables are dirt cheap.
The only reason to invest in expensive cables is if the length of the cable is so long that the standard cable fails to deliver enough output voltage at the receiving side. The only factor that matters is the damping per meter at the highest transmission frequency. As a hint: I have successfully tested 44/16 with 16 m of RG59.
The specs are more tolerant about impedance:
Fortunately, for a simple binary signal, there is no need to obey the transmission
line rules anything like as strictly as for an analogue video signal.
In fact the original 1983 specification allowed up to a 2:1 mis-match of the line characteristics and this gave a
certain flexibility to "loop through" receivers, or use multiple links radiating from transmitters. This concept
was based on the theory that lossy PVC analogue audio cables would be used and it was predicted that:
• reflections in short cables were unlikely to interfere with the edges of the signal, due to the short delays
involved,
• reflections in longer cables were likely to be attenuated so much that they would not significantly interfere
with the amplitude and shape of the signal at a receiver.
In practice, however it was soon found that problems occurred with an open ended spur which happened to
have an effective length of half a wavelength at the frequency of the "one" symbol. This length is also a
quarter wavelength for the frequency of the "zero" symbol. This condition causes the maximum trouble for the
signal characteristics on any connection in parallel with the spur.
It has been found that connectors are of little consequence since their electrical length is so short that any
reflections due to mis-match are immediately cancelled out. Surprisingly, some "noisy" analogue connectors,
such as brass ¼" jack plugs, work extremely well with digital interface signals. This is because all digital
signals, even silence, are still represented by several volts of data signal and, by analogue standards, digital
signals are very tolerant of crosstalk.
3.2. Guidelines for installation
The following practical guidelines have been produced for balanced circuits intended for the AES/EBU Digital
Audio interface. They are based on experience gained from two installations by the BBC in London and from
installations by CBC, Canada
Inter Area Cabling
- Cables should have a characteristic impedance of about 110 Ω (80-150 Ω is acceptable).