A CD can never achieve the fidelity of reproduction for quiet sounds and details of the highest possible quality of analogue recording.
It is not just the sample rate - as the others here have mentioned - it is also a case of the bit count.
The reason is that it only captures 16 bits of detail, and that is not nearly enough to capture the volume range of human hearing with a high level of fidelity.
Let's take the volume range of an orchestra as about 80 db from 30 db to occasional moments of 110 db (can be even more than that) at maximum volume (see STEREOPHONIC REPRODUCTION FROM FILM)
If you check up the definition of a Decibel then that corresponds to a volume ratio of about 10,000. But 2^13 = 8192.
So if we can match the recording level so that the highest volumes are captured with 16 bits of precision at exactly the volume they are played, then the lowest volumes in the orchestra are recorded with only 3 bits of precision.
In many recordings the loudest sounds aren't quite as loud as a full orchestra at maximum volume. If that happens, you will have a smaller volume range of say 50 decibels instead of 80, from say 30 db to 80 db. That still makes the quality of the quieter sounds about 7-8 bits.
Even 8 bits of precision is a poor quality of sound - I remember the old 8 bit recordings you used to get on the internet, and it is poor enough quality so you would only use it for music if you had no other option.
One reason CD recordings work is that most listeners aren't very sensitive to the sound quality of really quiet sounds, or have just got used to recordings that only capture the louder sounds faithfully.
It's not just the very quiet notes. It's also sounds that make up the quality of what you hear, e.g. guitar fret noise, or the sound of the bow pulled across the string for string instruments, small details you barely hear that make up the overall "live music" impression.
Here is an example, guitarist, notice how many quiet sounds you get in the recording from his fingers touching the strings
Also notice how much quieter these fret noises are than the sound of the instrument itself.
There are quiet sounds also in the individual notes too. The very beginning of the attack on each note, and the fade away at the end are quiet. Also of course the reverb of the notes is a quieter sound.
The Shackuhachi is an instrument where the sometimes very quiet breathy sounds you get as you play are very much part of the quality of the instrument sound,
Another reason why the poor sound quality of the quietest sounds is often not noticed is that commercial CDs are often normallised and compressed in volume range in order to make sure they sound good with only 16 bits of precision. See Dynamic range compression
So with CDs we have got used to hearing recordings with really low fidelity for the quietest sounds, and with many of the recordings compressed in volume range compared with a live performance (especially for music that typically has a wide dynamic range played live such as classical music).
That's why audiophiles recommend a much higher bit count such as 24 bits or 32 bits. With 32 bits, then those quietest sounds would have 19 bits of precision instead of just 3 bits of precision. They also recommend a faster sample rate as well.
Note on floating point
You also get Floating point formats such as 24 bit floating point digital audio. Also, often 32 bit audio is floating point.
With floating point you get a much larger dynamic range, but you get variations in quantization noise (precision and dynamic range) and a larger signal to noise ratio.
Also in floating point, if the quiet sounds and loud sounds are both played together, then the larger dynamic range of the floating point doesn't help. It only really helps if you have the quiet sounds played on their own.
So for both those reasons fixed point is generally better for audio recordings at 16 or 24 bit depth. Though when you get to 32 bit floating point then there are enough bits of precision so that floating point works fine, and it means you can mix recordings as you please without ever having any worries about hitting the dynamic range ceiling.
More links - Decibel (Loudness) Comparison Chart
Sound Advice Note 12 - representative sound levels of various instruments and full orchestra
It is not just the sample rate - as the others here have mentioned - it is also a case of the bit count.
The reason is that it only captures 16 bits of detail, and that is not nearly enough to capture the volume range of human hearing with a high level of fidelity.
Let's take the volume range of an orchestra as about 80 db from 30 db to occasional moments of 110 db (can be even more than that) at maximum volume (see STEREOPHONIC REPRODUCTION FROM FILM)
If you check up the definition of a Decibel then that corresponds to a volume ratio of about 10,000. But 2^13 = 8192.
So if we can match the recording level so that the highest volumes are captured with 16 bits of precision at exactly the volume they are played, then the lowest volumes in the orchestra are recorded with only 3 bits of precision.
In many recordings the loudest sounds aren't quite as loud as a full orchestra at maximum volume. If that happens, you will have a smaller volume range of say 50 decibels instead of 80, from say 30 db to 80 db. That still makes the quality of the quieter sounds about 7-8 bits.
Even 8 bits of precision is a poor quality of sound - I remember the old 8 bit recordings you used to get on the internet, and it is poor enough quality so you would only use it for music if you had no other option.
One reason CD recordings work is that most listeners aren't very sensitive to the sound quality of really quiet sounds, or have just got used to recordings that only capture the louder sounds faithfully.
It's not just the very quiet notes. It's also sounds that make up the quality of what you hear, e.g. guitar fret noise, or the sound of the bow pulled across the string for string instruments, small details you barely hear that make up the overall "live music" impression.
Here is an example, guitarist, notice how many quiet sounds you get in the recording from his fingers touching the strings
Also notice how much quieter these fret noises are than the sound of the instrument itself.
There are quiet sounds also in the individual notes too. The very beginning of the attack on each note, and the fade away at the end are quiet. Also of course the reverb of the notes is a quieter sound.
The Shackuhachi is an instrument where the sometimes very quiet breathy sounds you get as you play are very much part of the quality of the instrument sound,
Another reason why the poor sound quality of the quietest sounds is often not noticed is that commercial CDs are often normallised and compressed in volume range in order to make sure they sound good with only 16 bits of precision. See Dynamic range compression
So with CDs we have got used to hearing recordings with really low fidelity for the quietest sounds, and with many of the recordings compressed in volume range compared with a live performance (especially for music that typically has a wide dynamic range played live such as classical music).
That's why audiophiles recommend a much higher bit count such as 24 bits or 32 bits. With 32 bits, then those quietest sounds would have 19 bits of precision instead of just 3 bits of precision. They also recommend a faster sample rate as well.
Note on floating point
You also get Floating point formats such as 24 bit floating point digital audio. Also, often 32 bit audio is floating point.
With floating point you get a much larger dynamic range, but you get variations in quantization noise (precision and dynamic range) and a larger signal to noise ratio.
Also in floating point, if the quiet sounds and loud sounds are both played together, then the larger dynamic range of the floating point doesn't help. It only really helps if you have the quiet sounds played on their own.
So for both those reasons fixed point is generally better for audio recordings at 16 or 24 bit depth. Though when you get to 32 bit floating point then there are enough bits of precision so that floating point works fine, and it means you can mix recordings as you please without ever having any worries about hitting the dynamic range ceiling.
More links - Decibel (Loudness) Comparison Chart
Sound Advice Note 12 - representative sound levels of various instruments and full orchestra
About the Author
Robert Walker
Writer of articles on Mars and Space issues - Software Developer of Tune Smithy, Bounce Metronome etc.Studied at Wolfson College, Oxford
Lives in Isle of Mull
4.8m answer views110.4k this month
Top Writer2017, 2016, and 2015
Published WriterHuffPost, Slate, and 4 moreHuffPost, Slate, Forbes, Newsweek, Business Insider, and Science 2.0