It is understandable that little attention is usually paid to the quality of digital audio cabling. We are used to interconnecting our computer equipment with low-cost cables without mishap, and with digital audio it's rather logical to assume that no sound quality issues exist since we are simply moving digital data around.
But the choice of digital audio cabling can be important, because the problems of transmitting digital audio data aren't really the same as for computer data at all.
Data integrity issues
In general, digital audio interfacing problems are usually (but not always) the result of inadequate interface bandwidth, which is most often due to the choice of cabling. In extreme cases this can result in loss of data (and resulting dropouts in the audio) because (unlike computer interconnection protocols) simple digital audio interfaces such as AES3, S/PDIF and ADAT transmit the data only once, and without the possibility of error correction. Although there is a possibility that an error can be detected, this is of little use since no correction or retransmission is possible. So, unlike a computer interconnection, a mission-critical digital audio connection must ensure that no bit errors can EVER occur in the data stream EVER! This can be hard to guarantee in the real world, especially when the system sample rate is high.
This was not really much of a problem when these interfaces were first standardised, since the bandwidth requirement was quite modest when the maximum sample rate was only 48kHz. Unfortunately, back then, the use of analogue audio cables for digital audio transmission was actively encouraged by the choice of XLR and RCA/phono connectors for AES3 and S/PDIF respectively, even though they typically have poor bandwidth. But for AES3 and S/PDIF, the bandwidth requirement is directly proportional to the sample rate, since a fixed number of audio and status bits are transmitted per stereo sample (note that for ADAT/SMUX connections the bandwidth requirement does NOT rise with sample rate since the number of channels carried is reduced as the sample rate is increased instead).
Many modern digital audio devices can operate at sample rates as high as 192kHz, and (sad to say) many digital audio cabling setups don't have the bandwidth to support this reliably. Actually, it's worse than that - much of the 192kHz-capable equipment has digital audio ports which (either admittedly or otherwise) don't support reliable operation at 192kHz whatever cable is used. This is particularly true of TOSLINK ports (the optical variant of S/PDIF).
Conversion quality issues
But surely the sound quality of a digital audio setup can't depend on the choice of digital audio cabling, so long as all the data bits get through? Sadly, and familiarly, though - it can. Because in many cases the audio data stream is used to pass the sampling clock as well as the audio data between equipment. If the receiving equipment gets a clock which has been degraded by a low-bandwidth interface, and if it uses this clock for A/D or D/A conversion, then the sound quality of that box will be degraded. This effect is known as 'sampling jitter'. Unfortunately the biphase coding scheme used in AES3 and S/PDIF is very effective at converting low cable bandwidth into clock jitter. It should be pointed out that this is an entirely avoidable problem, since any box which relies on deriving a jitter-free clock for A/D or D/A conversion (or for sample-rate conversion) can take steps to eliminate incoming jitter - but many don't. The Prism Sound CleverClox technology in Atlas does exactly this, as explained in the Clocking and jitter section. This problem isn't really a cabling issue, but an equipment design issue. However, in most cases we can't change the design of poor-quality converters, but we can cover up their problems to some extent with good cabling!
Even though Atlas is insensitive to incoming clock jitter, and even though it transmits very low jitter at its digital audio and clock outputs, the question of cable quality may still be relevant if Atlas is transmitting to equipment which itself has poor jitter rejection capabilities. Note that audio quality degradation by cable-induced jitter is just as much a problem at low sample rates as at high sample rates.
A properly designed copper AES3 or S/PDIF interface will not cause audio-frequency ground continuity between the connected equipments, so hum loops should not occur. However, high-frequency ground continuity is essential if EMC legislation is to be met. This means that high-frequency interference such as from poor-quality switch-mode power supplies (see the Analogue interconnections section) can equally well be passed through copper digital audio interconnections. If this is a problem in your system, consider using a TOSLINK connection instead.
Maximising cable performance
In general, the best copper cable for digital audio is the cable with the lowest capacitance, since that will cause the least loss of bandwidth. For that reason, prefer cables specifically designed for digital audio, or for analogue video; don't use analogue audio cables - they don't have the bandwidth for digital audio use, especially at high sample rates. Prefer also the shortest cable, since (all other things being equal) loss of bandwidth is proportional to length.
Maximising cable bandwidth is important in optimising AES3 and S/PDIF data integrity at high sample rates such as 192kHz, and in optimising conversion quality in systems which include poor-quality converters. It is of little importance in protecting the data integrity of low sample rate systems, unless cable lengths are very long.
We are taught to choose cables of the correct impedance for the job. Whilst this doesn't have a direct impact on bandwidth, it can have a significant effect on data integrity at high sample rates and where cabling is short (and let's face it: at 192kHz cables had better be short...) because the reflections resulting from an impedance mismatch can affect the eye pattern at the receiver horribly. This can be much worse where non-matched connectors (such as XLRs) are present part way along the cable, such as in the case of 'breakout' cabled systems. For this reason, it may be better at high sample rates to use a continuous cable suitably terminated at each end rather than a 'breakout' arrangement.