I don't listen to CD's if I can help it--the technology of 16 bits starts distorting at about 4KHz. (Forget about Nyquist, that is for time invariant signals. A great paper out of MIT showed the math to support the distortion figures for CD's. BTW his conclusion was that 10X frequency is the minimum sampling rate--the same figure that Techtronix uses for the digital oscilliscopes). However, CD's can be made to sound MUCH better by the process of oversampling. What needs to be remembered is that oversampling does not add new original information to the signal. It does modifiy the 'jaggies' where the distortion is noticable by smoothing them in the convolution filter. This smoothing does add new information, but it does not necessarily duplicate the original sound.
To check this out a few years ago I did an interesting test. I took a good vinyl recording of the Four Seasons and recorded it digitally with pro gear at 44.1KHz, 48KHz, 88.2KHz, 96KHz all at 16 and 32 bits. Then I created a new 88.2 and 96KHz from the 44.1KHz by oversampling and re-masted it back down to 44.1KHz. I had this combo listened to by muscians (like myself), critical listeners, and people who by their own admission couldn't tell the difference between a tin cup and string vs. high end audio.
The results were somewhat surprising. First, the tin ear group couldn't really tell any difference between the recordings--not suprising and this was the group Sony listened to when making their decision about CD technology. The Critical listeners and Muscians all agreed that native vinyl was best, and that as frequency of sampling and word size went up--it got better and better. At 96KHz and 32 bit words, it was very close to native vinyl. Musicians could hear a difference, but the critical group didn't always. The pro studios typically use 196KHz sampling--and this makes sense given these results.
Where the surprise came was in the over-sampled recordings. Here the musicians all agreed that they sounded better than the 44.1KHz native, but that after listening to the original they stated that it just sounded wrong somehow. The critical group had no such problem--to them the over-sampling sounded great, even better than native sampling at a high speed.
I suspect that the difference is related to the fact that musicians, especially string muscians (like my listening group) hear the distortion in the harmonics and that sounds wrong. As a violinist, I learned to tune double and triple stops by listening for the harmonic convergence. This harmonic information is badly distorted in the over-sampling and filtering process. Critical listeners, however, have their ears trained by the recordings they listen to--not actual live music in most cases. Thus, they chose what they were trained by marketing to like.
Here is test you can do that will make a believer of you. Get your best CD recording of symphony that has lot's of cymbals in it. Listen until you are familiar. Then go to a live symphony and listen to the cymbals as they are played. You will swear it is not even the same instrument.
