Electronics notes/signal reflection
Shortish version
If you put an AC signal on a wire, it will be met with something that impedes it,
and when that effect is more complex than just the wire's resistance, we call that impedance.
Yes, that plain (DC) resistance is part of that, and in many situations the largest part.
Another ingredient is the result of things like the skin effect - and that will vary with frequency.
The implication one is that one wire is more fit to transmit a (higher) frequency than others
Also, there are reasons characteristic impedance might change somewhere in a path, which leads to some specific issues.
Consider
- the end of a wire, when there is nothing there, is effectively higher effective impedance than the rest (see below)
- physically different kinds of cable will have different impedance for the same-frequency signal
- (also, a short will be zero impedance)
When the impedance changes along a signal path changes, you see problems like
- signal reflection
- delayed versions of the signal (echo)
- attenuation (e.g. simple loss)
- attenuation distortion (different frequencies affected differently)
- signal ringing
- standing waves
Impedance matching means trying to make the whole signal path (all wiring) behave the same for the frequency you want, which avoids most such trouble.
When source and load devices have a designed purpose that is the same,
then they are often also impedance-matched, because some engineer did this for you.
In which case impedance matching comes down to using the right cable for the job.
Termination solves problems where the end of the wire's higher effective impedance causes of problem.
To minimize that effect, it helps a lot to have the end of the wire show a resistance that is the similar as the characteristic impedance of your wire with your given use. If your frequency use is well defined (it usually is) you can do this with a simple resistor.
Impedance mismatch is less of a noticeable problem on very short lines - up to a few meters you rarely have to care much at all.
Reflection only start mattering when the wavelength starts resembling real-world cable lengths. For sine waves this is a one or two MHz. For block waves (including most serial communication) it starts mattering a few factors earlier (e.g. in the form of ringing).
For example
- DMX's 250kHz rate, and the fact that applications easily use dozens of meter of wire, means it wants termination.
- It's not relevant for audio frequencies - even 20kHz (which is higher than most people can hear) is ~15km long.
Attempt at explanation
Most explanations just say that reflection happens, or give you a lot of words that associate it with various details, but which still come down to 'it happens'.
And yeah, you may want to choose to just trust that they do, because it is hard to get a thorough understanding of why things like reflections happen.
Analogy
The best analogy I found so far is rope. Not because it's accurate, but because it pokes your intuition.
You may remember a physics lesson (about standing waves and such) where you tied a rope to the wall, and give it a single whip. The point was that the wave comes back. Conservation of energy stuff, really.
Mixing impedances would be like combining thick/stiff and thin/loose rope, and giving that a whip. At the point where they connect, things happen that are no longer the wave you put in - a little energy goes back, the amplitude is different, etc.
If you're still swinging that rope when the result reflects back from the wall, things may get even messier.
Impedance matching would be not mixing rope.
Termination then would be something like a string damper on the end, which tries to dissipate the energy it gets so that it won't have to be hammered back. That damper's spring strength is critical, though - it's too stiff/loose for your particular hand-wave, it won't do much or even introduce trouble.
This is why terminators have a specific impedance.