There are some technical reasons for the four different techniques mentioned by SL1000. And there are misconceptions and there are myths.
As I see it, it started with audio amplifiers and microphones more than sixty years ago. Microphone signals were (and are) low level and the amplifiers had lots of amplification. So any "hum" picked up used to be very disturbing. Protective earth was not very common in those days and the result was hum pick up in the microphone cable. Screening helped to some extent, but if the screen was connected to earth at the microphone end, stray currents from amplifier through screen to earth induced hum in the cable. Isolating the far end of the screen helped a lot. That is how the first rule: "Connect screen in one end only" came about. And also how the dreaded "hum loop", which many people still refer to as if it was an eternal truth - valid in all installations. Which it is not.
Second scenery: Digital communication like RS232C and other crude techniques made its debut. These signals had immunity levels around a couple of volts and the "hum" was not really a problem any more. But transients and HF pollution were. Transients are HF, so the technique where the screen was grounded wherever possible started to be used. Many ground connections allowed transients and HF to find its way to ground without going through the input section of the DCT/DTE. So, "ground whenever you can" started to be a sucessful strategy.
The earlier communication lines were mostly intra-building and seldom more than twenty or thirty metres. Short-haul modems (remember the KM-1?) were used for longer distances.
When field buses made their entry, longer distances could be covered. Often between buildings. It then happened that the supply system for one building was not the same as for the other building. A quite common situation, actually. The result was that current was flowing in the screen if it was connected to the different grounds in the buildings. And if there was an earth fault in one building, the potential rise forced large currents through the screen and it very often scorched the screen and cable so that equipment failure resulted. That is when users started thinking.
The thinking resulted in different ways of solving the problem. There are simple means like grounding via HF capacitors, using carrier and isolating transformers, using opto-couplers, systems with enhanced common-mode range (like the RS485) and also bonding of the different buildings to a common potential. Every manufacturer and every communication standard now have found their preferred way of handling the problem with connecting screens and these installation are often very reliable. It is in special situations and when equipment from suppliers with different philosophies shall be connected together that problems start to pop up.
Those that haven't thought the problem through will then grab one of the earlier myths and try to apply the "avoid the dreaded hum loop" principle or the "ground everywhere" method. Sometimes both at the same time.
The best way of getting out of the mess is to analyse the situation and use the standard solutions that are available. Be it isolation amplifiers, fibre-optics, extended common-mode range, carrier frequeny and transformers, differential amplifiers or equipotential bonding throughout the plant.
But do not go with myths - and do not believe that old truths are always the only truth.
Gunnar Englund