It is in fact exactly the same as moving through the air; an induced magnetic field is normal to an electric field but otherwise basically two different appearances of the same phenomonon. Disregarding some implementation details like impedence and signal level, you could take the cable to a transmitting antenna and connect it to a receiver and get the signal out, skipping the airwave step entirely. (for a CATV system, this might be an especially handy way to think about it, since in many systems, the cable feed is literally the same kind of signals and the same frequency channels that would come in from your antenna)
Almost all transmissions are at their core analog; digital is simply a matter of how you encode and decode (doing so in a discrete manner rather than a continuous manner). Take a copper wire, hook up a bench power supply on one end and a multi-meter on the other in Vdc mode, and you can dial the voltage up and down and see that the voltage that comes out the other end more or less matches what you fed in (minus maybe a bit of loss, or +/- some error). Swap for a frequency generator and an oscilliscope, and you can feed a sine wave in one end and get mostly the same sine wave out the other end, again, with just a bit of error. You can dial up and down both the frequency and the amplitude (voltage), and within some frequency range if you leave the amplitude in the same, the amplitude out on your scope should stay the same too - this is more or less the usable frequency range for the cable (actually by convention, the BW is the width of the region bounded by the half-power points, but anyway).
Through Fourier transform / frequency-domain magic (greatly oversimplifying and hand-waving in order to avoid some truly nasty math), most signals can be modeled as a sum of a potentially infinite sequence of sine waves. That means that if you can get sine waves through in a mostly linear fashion and if you can model signals as a sum of sine waves, you can get most signals through. Different signals will have different base encodings (e.g., NTSC TV, and DOCSIS QAM), and through FDM they're simply modulated (for this discussion, shifted up in frequency) to the carrier channel they need to be on. All the different signals are simply summed together. On the receiver end, it's a matter of a demodulator that picks out a band of some necessary width and demodulates it (shifts it back down in frequency), then decodes it as appropriate.
DOCSIS last time I checked uses QAM for interesting data rates, which works by using both sine waves and cosine waves summed together - because the two are out of phase by just the right angle, you can put both on the same carrier frequency and encode and decode them independently, getting you about a two for one. But it gets better - you can use the (signed) amplitude of the sine and cosine components to build a grid. On that grid, put various bit patterns to be your code words. Take the amplitudes you have and find the nearest grid point, and there's your pattern. It's a very handy encoding.
Hopefully this makes things a bit clearer. In some sense you're asking for a large chunk of a EE program in a few paragraphs, but with some hand-waving hopefully I can give you a rough idea of how it works.
. Also learned about different digital and analog encodings (eg: Manchester, Pseudoternary, 8B6T, etc...). It's a fascinating subject but I just thought the prof went way to fast, and I don't understand it all of it at a level I wish. 
Facebook
Twitter