[. I prefer grounding only one end as that's an old audio trick for avoiding a ground loop, which could be completely irrelevant to this concept....
It was often wrong then and is even more likely to be wrong now.
And that old audio trick could cause a variety of problems in audio systems if you weren't careful. Connect to a device that has a broken ground plug but has decoupling caps from AC to ground and you could have 58.5VAC 60hz common mode; that isn't just hum, that could be a fried input. Connect to audio cables in series where one has the shield isolated on one end and you could end up with half your run unshielded - or none of it. Plug in a mic with phantom power into a cable with the shield broken and it isn't powered. Plug in a mic without phantom power into a shield broken cable and now the case of the mic is an antenna instead of ground. Shield isolation tricks like this, while they have their uses, are better done in a separate shield isolator that is used with care, where it is needed and only where it is needed, not done in audio cables themselves or indiscriminately.
Usually, it isn't the loops in the ground leads themselves (shields) that create "ground loops". It is when ground on one end and the other have different potentials on them due to other signals carried on one of them over a ground wire. Breaking the shield often doesn't break the ground loop, it can make it worse. But there are times when unwanted currents can flow through the shield and sometimes it really is advantageous to break the shield.
Also, at RF, the ground currents actually want to follow the path of the original current (due to mutual inductance with cancelling currents) instead of finding creative alternative paths if they can avoid it. Not connecting the shield is thus generally a very bad thing; it at least needs to be connected through a capacitor. Always give RF a return path that is right next to the original path; if you do anything other than that, it has a name: "antenna". Otherwise you are creating ground loops and EMI issues. In theory in a balanced line the other signal line is the return path but a dipole is not a fully balanced load in the real world.
For an antenna, the relevant grounding criteria is lightning protection not hum loops.
The shields should be grounded at all-three ends. What, you didn't know that coax had three ends?
- Top of tower
- bottom of tower. Not 3 feet from the bottom or 1 foot. Or else you get the lightning surge drop
across the inductance of those 1-3 feet of tower. You don't actually need to break the cut the coax
centers or shield here if you tie into shield (and waterproof). http://www.protectiongroup.com/ProtectionTechnologyGroup/media/PTG/WhitePapersandTechnicalNotes/1485-013.pdf?ext=.pdf
- entry panel into shack at lighting arrestor on single point ground plate
- fourth end: antenna tuner or transceiver. Normally a second piece of cable.
Better yet, don't follow a rule of thumb like "always break the shield at one end" or "always connect the shield at both ends" but instead look at the big picture for your specific system.