OK, JURB's basic electronics is open...
You do not need the 50 ohm impedance for the generator. It will put out a fairly square wave into any impedance that does not short it out. The termination for the cable is what's needed, not for the generator.
A square wave has harmonics (hopefully) up to like the 23rd or 25th. Take a 2 mHz square wave and multiply that up. Think 2 times 25 is 50. At that frequency, the speed of light comes into play. The scope has to absorb all the electrons put out by the source on a negative pulse and and supply all the electrons sucked by the source on the positive pulse. If this is not "matched" by the impedance, they clash, or crash, within the cable causing what can be referred to as standing waves. An example of this is the SWR meter used to match a transmitting antenna like on a CB radio. That SWR stands for "Standing Wave Ratio" and when there is too much it overloads the transmitter output and can blow the final (power output) amp.
In most audio work this is not going to bother you at all. The two most useful waveforms are the 1 kHz square ave and the 1 kHz (or 400 Hz) sine wave. Used for two different purposes, but really, neither requires dealing with RF. The 23rd harmonic of 1 kHz is of course 23 kHz. This is no problem.
If you look up "poor mans tdr" on youtube there is a good video of a guy who makes a super fast rise time generator to do actual time domain reflectometry. I cannot supply the link right now because of a PC problem, damnthing has Vista and every once in a while youtube screws up my sound drivers and I have to reboot. Linux is coming soon, believe me. He built the generator out of Schmitt triggers. I think the IC has six and he turns one into an oscillator and uses all the rest of them in parallel through five 200 ohm resistors to get the correct output impedance. The scope is the 50 ohm terminator. Then he feeds it to a cable and actually measures the nano or picoseconds on the screen and derives the length of the cable from the reflection indicated on the scope. I think it was like a ¼" error and he explained that was due to the connector itself.
I looked back in my email for this, no find. There is a ten minute video that looks like it in a search, but I am not sure if that is it. Try it :
https://www.youtube.com/watch?v=Il_eju4D_TM
And my downloader isn't friggin working or I would Dropbox it.
At any rate, think of water flowing through a pipe. You want the pipe to be exactly half full, and for the output to accept the water at exactly the same rate as it is being put in the other end. As long as this is the case, the flow is nice and smooth and there are no waves. If the receiving end stops up a bit and accepts less than the exact amount the pump is putting out, there will be waves. If the receiving end dumps the water out faster than the other pump pumps it in, there will be waves. You do not want the waves. Mechanically such a situation wastes power (in electronics also) but in electronics it distorts waveforms with echoes, the combing effect and phase shift.
That is the reason for termination. To keep from making waves inside the wire.
I hope this helped at least a little toward understanding about impedance matching on a cable. I can't come up with a better "laymen's" way of explaining it right now. I really do not understand it completely myself. I do enough to use it and all that but there is a bunch opf complicated math involved that I just don't get.
For example the cable impedance. It is based on the resistivity of the conductors and their distance apart, from shield to the center wire of coaxial cable. When you have a 50 ohm source and a 50 ohm terminated destination, you must have a 50 ohm cable or it does not work right. (the pipe is the wrong size)
I was trying to align an HP 1725a a while back and it just didn't look right. I used my Wavetek 111 50 ohm output and had the HP set to terminate 50 ohms. Ringing all over the place ! WTF ! I had another look at the cable and see it is RG59U, which is 75 ohms. Those cables were bought surplus, and I imagine they were used for audio or something and the RG59 was cheap so they used it, for that one purpose. Either it was low frequency and it didn't matter or it was designed for 75 ohms but happened to use BNC connectors. If there IS an impedance involved, BNC connectors are for 50 ohms. Those cables at the back of your cable box are 75 ohms, and the connectors are made to keep that impedance as well as possible. So even if I changed the impedance of the system to 75 ohms, those connectors would throw it off a bit.
That RF is funny stuff. you can't really switch it all that well without going to a design from Mars or something. And it is part of your square wave.
But really, now that I have orated and you all owe me tuition or a bullet in the head, to answer your question - no. You do not HAVE to terminate it unless you are getting into the higher frequencies. For audio, most applications you can just cut the cord off a toaster and it will work just fine.
