Ah shit. This is tricky.
Right.
I'll assume that you know that it is possible to record at 16bit, 24bit and 32bit?
well, this refers to the accuracy of the recording. When you digitally record a sample, it has to give a number to each segment of sound it records, normally an integer.
The thing is, digital recording is only ever an approximation of an actual sound source, as it is not a continuous recording method. It operates a lot like a flick book does for animation. The idea is, that if you flick through the images fast enough, you'll never notice that they are still.
This means, that there needs to be a method to determine what integer is produced.
At 16 bit accuracy, that means that there are 9999999999999999 ( I think ) possible amplitudes ( think of waveforms hear). 'Great', I hear you say, 'surely that is waaaaay more then enough possibilities to record every single musical nuance ever?', and in theory if everything is recorded at zero db, then yes it is. The problem is, as the source is recorded more quietly, less integers are used, effectively reducing the bit rate.
Get out a bit crusher and apply it to some audio. Open up a waveform editor, and compare it to the original. Notice how the bit crushed waveform is full of straight lines, that is becuase you have reduced the resolution of the recording effectively. Now, zoom in on the original waveform, allllllll the way. Guess, what that is right it is full of straight lines, just much closer together.
The effect that a bit crusher has, in very unsubtle terms is quantization noise.
What is the advantage of 24 bit audio then? Well, you have 999999999999999999999999 possible integer values in 24 bit audio. What this means is that you can record things with a much greater degree accuracy, but still when things get, really quiet, then you will still get this reduced bit rate effect, otherwise known as quantisation noise, as it simple runs out of numbers.
32 bit audio is different. It doesn't deal with integers. It actually has the same recording resolution as 24 bit audio, with one vital difference, it can use the decimal point. when a sample is only using say the first five integers of amplitude, say 99999, it will put in a decimal place and capture the reamining 19 integers of detail.
In practice, what this means is that recording resolution stays the same at much greater variety of volume, with no bit crushing effect.
Hope this is clear, apologies for any errors I may have put across.
