Free-space versus Real-world Dimensions
The term "dipole" simply refers to an object having "two poles". There is no particular size associated with a dipole. It can be made from two thin wires, two spherical balls, or even two parallel square plates. One can observe radiation of a dipole of any size. Now it just so happens that if a dipole is made from thin wire or tubing of total length 468 feet/ f (MHz), or nearly a half wavelength long, and the wire is cut to insert a feedpoint, and the wire dipole is placed at least a quarter wavelength away from other objects, then we can observe an approximately resistive impedance at the dipole feedpoint. It is observed to be "resonant". The resistance we observe depends on where we insert the feedpoint. If the feedpoint is in the center, we might see about 70 ohms. Off-center the feed by 0.12 wavelength, and we might see 200 ohms. If the feedpoint is almost at one end, we might see 2000 ohms. All are perfectly good resistors, just different values. Speakers are often 8 ohms, and they work fine. The goal here is too get a resistive load that can take power with a resistance value of something the power amplifier in the transmitter can drive to rated power. Most ham transmitters are specified to put out, say, 100 watts into a 50 ohm resistor. Since a "center-fed halfwave dipole" presents about 70 ohms at its feedpoint, we could probably use that antenna directly for a load for the transmitter. And we do.
Dipole Performance
Several configurations of the wire dipole antenna are suggested in Figure 2-1. The wire dipoles are all hung mostly horizontally. It should be pointed out that these configurations all radiate a good fraction of their input power broadside to the plane of the wire and horizontally polarized. If the antenna is near the earth (ground), then in the direction of the wire they radiate mostly vertically polarized signals. This is true for the very common inverted vee configuration; for an antenna hung about a quarter-wave up, at 10 elevation takeoff angle, broadside is horizontal polarization, while off the ends is vertical polarization, at almost equal power. That is a very good all-purpose configuration.
Dipole Construction
I agree. Stranded wire is best. Copperweld (copper-clad steel wire) is stronger if it doesn't rust through. I myself use regular solid insulated housewire for inverted vees or short dipoles. It's usually quite adequate.
Multiband Dipoles
Since any resistance value is as good as any other and series inductance can be resonated and tuned out by series capacitance, and vice versa, we can let the dipole impedance be almost anything ... if we use a Transmatch or antenna tuner to transform the antenna and feedline to 50 ohms for the transmitter to drive. That means almost any old length for a dipole through any length of feedline will work. Well ... usually. It helps if the dipole is at least a quarter wave total length. It helps if the feedline is extremely low-loss, like open wire line. It helps if the characteristic impedance of the feedline is somewhere around 300-600 ohms. It helps if the antenna tuner and baluns are low-loss with the load they see. But these caveats just help minimze the losses and maximize the probability that the load for the antenna tuner is one that it can transform to 50 ohms with some settings.
The G5RV antenna is a 102 foot dipole, center-fed with 450 ohm line. That makes it 5/8 wavelength long on each side on 20 meters. Here it has a slightly squished pattern and an extra 3 dB of gain. You need a tuner to match into it, but for just 20 meters, you could find a good length of 450 ohm feedline (about 30 feet) that happens to transform the antenna impedance close to 50 ohms. Several other hambands also seem to transform pretty well with that same feedline. Using it higher than 15 MHz causes the antenna pattern "break up", to have many lobes and nulls, so you could have a null right where you want coverage. But for casual operation, it's very popular.
Fanned Multiband Dipoles
Yes, you can put a bunch of dipoles, each cut for a particular frequency band, all in parallel on the same feedline. If the dipoles are spread w-i-d-e apart like a fan, the tuning of one won't interact much with the tuning of another. The idea is that all unused dipoles are pretty much high impedance while the one that's in use is near 70 ohms ... and they're all in parallel, so the result is 70 ohms and the rig is "happy". This is not a bad solution.
Trap Dipoles
Doug discussion of trap dipoles covers almost all the bases. Trap dipoles consist of one center-fed dipole wire with "traps" (parallel L-C's) strategically placed along it. The operation is simple. Say we wanted to make a trap dipole for 10 and 15 meters. What we do is cut a dipole for 10 meters, about 16 feet long. At each end we place a 10 meter trap, a parallel L-C tuned to 28 MHz. Now anything we add to the other side of that parallel L-C will be "disconnected" around 28 MHz. So the 10 meter dipole should work just fine. For 15 meter operation, the L-C trap is far below resonance and just acts like a small inductor. We can add some more wire beyond each trap then to bring our antenna to resonance at 21 MHz too. Since the trap has that small inductance, we won't wind up with a 22 foot dipole, but it will be a little shorter. That's it. We can do more bands just by adding more traps to disconnect the sections of wire beyond them. Neat, huh?
Dipole Look-alikes
Doug mentions long wire antennas (length > 1 w.l.) andcollinear arrays, but we've seen this type even on the G5RV at higher frequency.
Added Notes