Fading - A phenomenon I encountered when I first began experimenting with radios in the 80s. It occurs when a radio’s signal strength varies with small movements. Walk around with a handheld radio or listen to a car radio; the received signal strength will vary greatly. Sometimes the fluctuations are random and other times peaks and troughs occur every half wavelength travelled. 
While living in Beaty Towers dorm at the University of Florida in Gainesville, I experienced a striking case of fading while talking to talking to a friend 120 miles away through a 2m-band [146MHz] repeater located on the roof of my building. (You can see the antenna at the picture to the right from floridamemory.com, which was taken around the time I lived there.) The station in Clearwater could reach the repeater with 50 watts into a antenna with 22dBi gain. I reached the repeater from my fourth-floor dorm room with 50mW into a rubber duck. There were some locations in the building where I could not reach the repeater with my transmitter. It was shocking to observe this happen while talking to someone over 100 miles away who was getting into the repeater with no difficulty. I walked around the building and imagined all there places where reflections could occur.
In my communications class at UF I learned complicated models for this type of fading. The case in which the transmitter and receiver have a direct path between them plus some reflections is modeled with a Rician distribution. When there is no direct path, the signal strength tends to vary more, and this is modeled a Rayleigh distribution. When I first heard this, I thought academics must be scraping the bottom of the barrel for things to model in MATLAB and write transaction papers about because to me fading was just a fact of life. These models surprisingly turned out the basis for the most amazing technology I have worked on.
The most basic use for modeling a fading channel is to work out how reliable a radio system will be. If the receiver requires -80dBm of signal to work and you calculate the signal strength will be -70dBm, you can use a fading model to work out what percentage of the time the signal will fall below the -80dBm threshold.
You can also work out when the reflections will be far enough apart that by the time the later reflections reach the receiver the message on the signal has changed. For example, a typical speech sound (phoneme) lasts 20ms to 40ms. Radio waves propagate a 5us per mile, so if you had a radio system transmitting voice with a strong reflections taking paths 4000 miles apart in length, the reflected signal would contain sounds 20ms delays from the other, making it difficult to understand. You typically only see reflections from a few miles, so this intersymbol interference is not a problem for voice signals.
A Wi-Fi (802.11a/g) signal sends symbols that last only 4us. Reflections from paths one mile apart (i.e. 5us of delay spread) are an issue that people setting up Wi-Fi based systems occasionally encounter. Reflections significantly shorter than the symbol period, cause the signal strength to vary but don’t corrupt the data.
In school I saw transaction papers modelling how having two antennas receiving the same signal could reduce fading because both antennas being in a location with low signal strength is much rarer than a single antenna being in a bad spot. They took a step further and said a multiple input multiple output (MIMO) system could have multiple transmit antennas. The receiver could use the variations in echoes received on its two antennas to receive signals that would be too weak to be received on a signal antenna. The transmitter could theoretically send completely different data on its multiple antennas on the same frequency. The receiver could use the different delay spreads of the paths between each transmitter and receiver to recover multiple streams of data. When I read that in 2002 I thought this was MATLAB fantasy.
Seven years later I worked on a project with 802.11n, which uses multi-stream MIMO. 802.11n is pretty finicky about the channel. It needs not too much or too little delay spread to work in multi-stream mode. It works best if the antennas have different polarizations. But I have seen it realized and work well on a $50 mini-PCI card.
These random fluctuations in signal path that I originally thought of as nothing more than a random fluttering of leaf in the wind became the basis for the most amazing technology I’ve worked with. If you had told me about it in the 80s when I had a 2400 baud phone modem, I might have been able to imagine sending megabits per second over a wireless link. I certainly would not have believed you could transmit multiple streams of data, each one in the 50Mbps, range and recover them based on the antennas being 100 light-nanoseconds apart.
I don’t scoff at predictions about curing aging or technological singularity. The most mundane phenomena can lead to revolutionary technologies.