The term "analog," when used in an audio context, refers to a signal that directly represents the sound waves traveling through the air. A simple tone, such as a sine wave, causes the air to form evenly spaced ripples of alternating high and low pressure. When these waves reach an eardrum, or a microphone, they in turn cause it to move evenly back and forth at the same rate. If you measure the voltage coming from the microphone and plot the change over time, it looks like this:
The earliest forms of audio storage media were mechanical devices that carved the motions of a vibrating membrane directly onto a wax cylinder. The grooves could then be "read" by a stylus and amplified through a megaphone-like loudspeaker. As the technology advanced, the cylinders were replaced by discs, and the some of the mechanical parts were replaced by electronics. The apex of this evolution was magnetic tape, wherein the signal was stored as levels of magnetic intensity, or flux, on a long piece of plastic with a coating of iron oxide particles. The neatest parts of this scheme were the ability to record over a mistake or add additional tracks, and the ability to splice pieces of tape together to edit the material.But then computers became important, and in order to deal with information, a computer has to have it spoon-fed in binary form (zeroes and ones). So, if audio is going to be stored, edited, manipulated or transmitted by computers, it must be digitized first. There are several methods, but the end result is pretty much the same - incoming voltage levels are converted into binary numbers. This is done with two important constraints: sampling frequency (how often the voltage is measured) and resolution (the size of numbers used to measure the voltage). If we take the sine wave pictured above and digitize it, we are essentially taking "snapshots" of the waveform at a certain interval, called the sampling frequency, and storing those snapshots as binary numbers one after another. Now, if we reconstruct the waveform from our list of numbers, we will see a "stairstep" approximation of what we started with:
When this is converted back into voltages, the steps are reconnected with lines, and the end result looks and sounds like the original - if the steps are close enough together. If the sampling frequency, resolution, or both are too low, the reconstructed waveform will be of lower quality.
