Quote Originally Posted by woodman
I'd be better able to explain this to you if I only knew how to draw diagrams directly into the computer which I don't. My computer literacy is not what I wish it were. Having said that, let me try to make this concept understandable to you.

An analog audio waveform is comprised of a multitude of up and down "squiggles" (for want of a better term). These up and down excursions are constantly varying in both speed (frequency) and amplitude (volume).

In order to convert this analog waveform into the digital format, the waveform is "sampled" by an A to D converter at TWICE the frequency of the highest pitched sound that a given format is capable of. In the conventional "Redbook" CD format, this "sampling rate" is 44.1 thousand times per second, which means that the format can store and reproduce any frquency up to 22KHz (22 thousand "cycles" per second which is higher than any human that I know can hear). This sampling yields a voltage value of the waveform at that instant in time. It is translated into a 16 bit "word" of 1s and 0s which can represent any of a possible 65,536 different values.

In order to convert this digital signal back into analog form (so that it can be heard as sound), a D to A converter is used that takes each digital "word" sample and converts it to a voltage value that is theoretically the exact same value as the original analog signal voltage that the sample represents. Therefore, it's possible to reconstruct the original audio waveform to its original state - recreating the original sound in (almost) perfect fashion.

I hope this made the whole concept a bit clearer for you
If I have understood what you are saying there are 65,535 different possible values a given point on a waveform can have within a range of 0(?) - 22,000 Hz on a CD. This implies to me that the accuracy of the representation of that waveform is limited.

This would mean that for 96KHz with a 16 bit word the accuracy of the representation of the waveform would be worse as the range is now 96,000/2 or 48 KHz to be represented with the same 65,535 possible values.

However if we use a 24 bit word (with 16,777,215 possible values) the accuracy of the representation of the waveform is higher.

Correct? Or am I a mile off base again?