A more complex approach to lightning + surge protection...
it's interesting that noone has stressed the need for complex lightning protection, at least noone has gone into deeper details. I am no expert myself, but I do know the basics, because I do deal with antenna stuff marginally, both on the job and as a hobby. Let me explain a bit about how lightning protection is dealt with here where I live = in Europe, across the pond.
Call us over-regulated, but we do have actual *technical norms* regulating how lightning protection should be built - mandatory for newly built buildings, and a good idea for any other house. I believe the number is EN 62305 or IEC 62305. Did someone mention Dehn und Sohne, the German vendor of lightning-protection gear and surge arrestors? I believe they're a major contributor to the norm. Even if this is a bit of a vested interest, I'm glad to have Dehn research this stuff.
The concept of complex (literally "coordinated") lightning protection of a building breaks the problem down into two parts:
1) outside lightning protection (the rods and collection conductors, plus their grounding)
2) surge protection for sensitive equipment inside.
For the outside protection, the norm specifies what the outside collection system should look like - how dense a network of down-conductors / collectors, how thick conductors to use, how many collection rods... for instance, the down-conductors should be spaced no further apart than 10 m (I believe), which means that a mid-size family house will have four, rather than just two, down-conductors.
This fairly recent norm is pretty pedantic and quite demanding engineering-wise. It specifies some spatial models of the protected space, which (some say) require the engineer / system designer to use specialized software, to be able to model/verify the building and its outside protection system thoroughly, to make sure the building is completely covered. Ever heard of the "rolling ball model", with 20 or 40m diameter, corresponding to different "statistical confidence levels"? In the old norms, it used to be a simple cone, spreading from the collection rod...
An important part of the outside lightning protection system is a good solid ground/earth. I hope I'm not wrong to say that again newly built houses must have a stainless steel grounding stripe (of a specified cross-section, I believe 4x30 mm) buried in the concrete of the foundations (below the watertight insulation) around the whole perimeter of the house - and a certified "electric mains revision engineer" has recently suggested to me to lay a dozen meters of this stripe into the ditch I was digging for my incoming mains service cable. The stripe buried in the wet concrete must have an appropriate number of "attachment terminals" clamped on, corresponding to the number of lightning collector down-conductors coming off the roof. When I was a lad, my father used to tell me that a good earth can be a square meter of a stainless steel sheet, buried deep enough in the ground to stay in wet contact with the earth all year long => noone ever told me that a simple rod rammed into the ground would cut it.
As for inside surge protection: the building and its closest surrounding open space is "compartmented" into so called "lightning protection zones": LPZ 0 for outside, LPZ 1 is the first zone inside the building, LPZ 2 is a "machine room" deeper in the building. And, on any cabling entrances / interfaces between the zones, there shall be surge arrestors of appropriate "damage/energy capacity", nominally called "protection class". Until a few years back, the classes were:
LPZ0 -> LPZ1 = class B lightning current arrestor (at the entrance into the building)
LPZ1 -> LPZ2 = class C surge arrestor (at the entrance into the room)
and there are also class D surge arrestors, inside sensitive equipment.
Nowadays the classes were renamed to "1, 2 and 3", probably to make the matter more confusing
The stratification of surge arrestors into class B, C and D has two aspects:
1) energy of the pulse - typically stated as a peak current in kA, over a predefined wave in time:10/350us for class B, 8/20 us for class C (the two time values being rise time and fall time to half the peak value).
2) response time of the protection device - the finer stages in the cascade should react faster.
And again, the building has an inside protection earth - the green/yellow conductor. For many years now, our mains distribution system (in newly constructed buildings) has a dedicated protection earth, separate from the neutral wire. Apart from other benefits, it allows for deployment of "current-compensated" protection breakers that sense undesired leakage from the live+neutral wires. Back to the point though: the PE and N conductors are only connected together at a single point in the mains service cabinet of the building. And, this inside earth is connected to the aforementioned outside "lightning protection earth", again at a single point - on the sketches, this point is typically painted someplace on the underside of the building The point is: if you get a direct hit from a lightning, the grounding system (outer and inner) should make sure that at least all the protective grounding outside and inside remains bolted together at some common potential - I believe hence the label "equipotential grounding busbar" used often in our mains and protection literature. Combined with a functional outer lightning collection system, this should make sure that noone inside the building gets hit by a voltage spike.
Just a note on the surge arrestors: obviously there are different arrestor gadgets for different purposes, using different protective devices on the inside: there are protection gadgets for the mains line, for telephone lines, twisted-pair Ethernet, for small-signal coaxial cabling, for various industrial purposes (slow analog signals, fieldbusses etc). As for the bare electronic components used: perhaps the most universal component is a "gas-filled static discharge chamber". These can be used up to several GHz as they have very little parasitic capacity - but their opening voltage is at least 90 V, which is failry high for small-signal electronics. I.e. the gas discharge capsule is a fairly coarse protection. For finer protection, there are zeners and especially transils = semiconductor devices, with nominal opening voltages starting from units of Volts. Their major downside is a significant parasitic capacity, making them useless above perhaps 10 MHz - another downside is quite some leakage current well below the opening voltage. For signals below maybe 200 mV, you can even use plain Si or Schottky diodes - again beware of parasitic capacity (matters for RF signals). The beefier semiconductor you choose, the more parasitic capacity. Coaxial transmission lines are nice in that the outside conductor (coax shielding) can often be directly grounded = pretty good protection. Obviously you still have to protect the live conductor with some arrestor device. For relatively narrow-band radio signals (anything modulated on some RF carrier) there is a perfect protection device: the quarter-wavelength shorted stub. For DC (and for lightning), this is effectively a short - but it passes the RF energy at its center frequency as if it wasn't there. It also has a downside: it's incompatible with phantom DC power, often used to power an LNB / downconverter / preamp in the antenna. Thus, for broadband / multi-frequency RF above say 10 MHz, your only option is the gas discharge capsule. Still the coax seems a lot easier to protect than free-floating symmetrical transmission lines, such as Ethernet or the telephone line. The "protection gadgets" on the market, tailor-made for a specific purpose, often contain further devices, apart from the surge arrestor component itself - there can be a combination of several arrestor devices (depending on the arrangement of live signal wires and reference ground, number of pairs), the arrestor devices can be in a cascade, and can also be supplemented by R/L/C components (filter) to improve the protection effect of the gadget. The effectiveness of serial resistance or inductance is only limited by the required passband (again this is a problem with RF). For some signals, galvanic isolation (on the "safer side" after some surge arrestors) can be a good additional measure to increase protection.
An important point: you have to protect *all* the cables/lines entering the protected space (house), otherwise the protection isn't complete. And yes it's impossible to protect signal cabling against a direct lightning strike. Which doesn't mean you should give up entirely.
As far as antennas on the roof are concerned, these should be placed in LPZ 0b (= shaded by the lightning rods), and the new norm demands not only galvanic isolation of the antenna from the outside lightning collector and rods, it further specifies that the isolation should fulfill a so called "safe distance", which is something like 1m of free air. If you have an iron mast for your antennas, that mast should be that well isolated from the collector. The recommended way to do it is not to ground the mast to the lightning collector, and use some plastic supports (1m poles) to mount a tall lightning rod high enough over your antennas. Obviously its downconductor should keep the required distance. See a photo in the references.
Finally, let me offer a couple of links to sketches and further reading. Especially, the nice colourful sketches in the Dehn publications are much more useful than the spartan simple black-and-white sketches in the official norm (the norm deals with measurement/evaluation methods for certification purposes, and the relevant sketches are quite narrowly focused / theoretical).
http://www.dehn.de/pdf/ds/DS614e.pdf (page 6, zoom in on the sketch)