RSI -- Welcome to reality. I design audio electronics, so I can answer your question.
It starts with the fact that CD's suck compared to old analog recordings, but we'll get to that at the end of this piece. As you know, CD recordings contain data -- sixteen bits @ 44 KHz x 2 channels, plus other identifying info, such as the title and location of tunes on the disk and control and formatting information. A complete CD holds about 650 MB. It's not exactly accurate, but if there are ten tunes on the disk, then about 65 MB would be enough for one tune. There simply is not enough bandwidth on a 56K line to transfer an entire CD and play it in any kind of useful time.
Enter data compression. Unlike PKZIP, which is a lossless compression scheme that alloww you to retrieve all of the original information in the file, MP-3, Dolby AC-3 and other, similar formats use lossy compression. In these schemes, a computer uses principles of psychoacoustic masking eliminates parts of the sound it "thinks" you can't hear under the louder parts of the sound.
Here's a clue. I belong to a couple of professional audio groups. At one meeting, we ran a CD through a Dolby encoder and digital delay that gave 100% accurate data with enough delay to match the output from the encoder in time. We then subtracted one output from the other. What remained was just the parts that the computer removed. Every person in that room who was experienced in recording was shocked. What we heard was a bunch of subtle subtextures, such as room echos and other audible cues that no experienced recordist would want left out of their carefully crafted mix. In other words, such schemes will never give you an exact copy of your CD.
As I said at the start, CD's suck, too, compared to original sounds. The sampling rate is way too low, and there just aren't enough bits. The inherant distortion in CD's is non harmonic. That means, unlike harmonic distortion (THD), the distortion products are out of tune with the music, which, in turn, means that human beings are far more sensitive to this kind of distortion. If that wasn't enough, unlike almost every musical sound generator, amplifier and speaker, the distortion gets worse as the music gets softer. Therefore, when it's full bore blowing your ears into distortion, it's as clean as it's going to get. In a moderately soft passage, where your ears are more sensitive to distortion, CD's are glad to give you lots more distortion.
44 KHz is an inadequate sample rate. This sampling rate was chosen based on Nyquist's theorem, which states that, to recover a given frequency, you must sample the information slightly more than twice the highest frequency. The problem is that, as you get closer to the high end of the audio spectrum, this theorem is only valid for a single, steady state tone. If you change the conditions to allow for a second tone, or to modulate the amplitude (volume) of the sine wave while it is being sampled, you have created a condition where there are literally an infinite number of possible outputs for a given sample.
There is hope on the horizon. The highest standard for the new audio only DVD is two channels of 24 bit data @ 192 KHz with only lossless compression. At that sampling rate, it will once again matter if the analog electronics I design can do a good job of reproducing the signal.
Don't worry. It's a multi-format standard that is compatible back to current CD's, so you'll still be able to play them. Of course, once you hear the new stuff on a good system, you may not want to, anymore. We may finally be about to come out of the Audio Dark Ages[b/].