A few points from over 30 years in industrial controls and assorted observed issues with pairs,shielding and grounding for a variety of signal types. 1) The "l

A few points from over 30 years in industrial controls and assorted observed issues with pairs,shielding and grounding for a variety of signal types. 1) The "less than 1V between buildings". Unless the buildings are purposefully and directly bonded together, don't count on it, seen 12 volts at 50 metres, various reasons. However, the design criteria for successful and robust systems of this nature is rarely defined by performance during expected conditions, rather by performance during fault or surge conditions. eg electrical fault in a nearby sub or a lightning event. In which case, unless direct and intentional bonding of some substance is installed between the buildings, forget about your 1V, try 100V, or even 10,000V in extreme cases. The golden rule of earthing and bonding - bond the fuck out of it, all boats rise in a rising tide, it is the potential difference that gets you, not that absolute. 2) The whole shield/core as an antenna thing as described here is basically correct, in theory. But, as soon as you have even a single installation (or maybe design/spec) problem, you could have an antenna and possible induced currents introduced to your system, with unpredictable system problems. Where there might be isolation by tx or opto, you have extra layers of protection. But as soon as you have one or more unterminated cores (eg spares left floating, broken wire etc etc) you now potentially have lost your golden ticket, being the cancellation of induced voltages via a changing B field, by courtesy of the balanced pairs being twisted. But have no doubts, if you have a rare case and get unlucky enough to have a resonant length of any kind of conductor/wire unearthed at one end (typically quarter wave resonance is enough, but any multiple of) and a matching noise frequency of any substance, you can have problems. The amount of isolation given by the use of very high dielectric core insulation (to allow such high speeds of data transmission without capacitive rounding of edges) means there is possibility for resonance with a high Q factor, meaning induced voltages can be much, much higher than field strength calcs that don't consider Q factor would indicate, and cause very real and serious problems. I have seen this effect act in ways that you would swear were magic if you did not see them. I had one where a hand held two way radio, in exactly the right position, seemingly using pipework as waveguide for over 30 metres, induced a voltage in a Siemens actuator that swung a 1 metre diameter butterfly valve closed. It was controlling pressure in a hydrogen disposal furnace running at 1000 m3 an hour, for unhappy results. No related wires for 20+ metres from the radio. Yet you could go stand next to the damper actuator and key the radio all day for no effect. Even after detailed documentation and video recordings etc were provided, Siemens refused to believe it was a real problem, in equipment certified for use in gas service. Resonance problems are rare, but when you get one it can literally cause problems that look like a hex or magic is happening. 3) There is also a whole seperate fact sheet or paper that should be read as well potentially, that being regarding use of terminating resistances and how/why/when for regular and special cases, but at GBs frequencies the wavelengths are so short so as not to be of concern, but this could be relevant for induced noise. 4) The game has changed for industrial systems in the last 15 or so years (primarily in 4-20mA current loops, but also some comms and serial links) from rule of thumb of earthing shield at one end only, to earthing at both ends. The reason? The noise spectrum has changed to include much more HF noise, up to at least 10Mhz, whereas previously you might have mainly worried about 3rd harmonics of your power frequency, so 150 or 180 Hz. The main culprit? Constant current power supplies, AKA led lighting. It's everywhere now and many are poorly designed and shielded and totally noisy as fuck, it's a fucking nightmare. Some installations I have seen, night time comes and the odd thing starts behaving a little strange. It's like black magic and the witching hour is in the air, but no, it's your lighting, and the haphazard cable installs for it, because "it's only lighting circuits". Just a few gotchas to look out for, I have endless far more nuanced details and bizzare faults accumulated over the years - "just stand there holding that cable and don't move while we get a download from the analyser to the PLC for this hours assays" and "OK, who stood on the cable in the tray because now we got a reflection and it's dropping data", when you can barely see a mark on the cable. As I said, the true quality of a design/installation in this area is system performance under abnormal/fault or unexpected conditions. There can be a lot of very subtle details that are only understood the hard way.