Or, “Why 24-bit audio is a load of shit”
I read yesterday that Apple is planning to start selling 24-bit songs on iTunes, and this made me angry. iTunes currently sells 16-bit 256 kb/s AAC files, and will start selling 24-bit 256 kb/s AAC files.
Nobody will be able to tell the difference. Not a single person.
I say this as an audiophile whose entire music collection is stored in lossless FLAC files, because I claim to be able to tell the difference between FLACs and 320 kb/s MP3s. 24-bit is not going to make any difference.
So, let me back up my pessimism with some theory first.
Let’s say you have a 24-bit audio signal. Every sample in the signal is a value between 0 and 2^24 which represents the amplitude of the signal at that point in time. If you were to convert the entire signal to 16-bit, the lowest 8 bits of every sample would be lost — rounded, truncated, whatever. Whatever the true value in the original 24-bit signal was, the 16-bit converted sample will be no more than 2^8 off from the true sample — the difference between the true value and the downsampled value can be no greater than 2^8. This difference of 2^8 represents, at most, 1/65536th of the value of the original sample (2^24 / 2^8 = 2^16, or 65536). This is a difference of about 48 decibels. To put that into perspective, if you played a 16-bit signal at an equivalent volume to rainfall, the lost information in that signal would have the volume of the softest sound your ears could hear if there were complete silence otherwise.
So, that was the theory. Let me also back up my claim with an experiment.
Here is a 5-second clip from a 24-bit FLAC song that I happen to have in my collection. (The song is Progenies of the Great Apocalypse by Dimmu Borgir.)
Here is the same 5-second clip, but downsampled to 16-bit using Audacity. (Spoiler, it sounds the same).
I loaded both of them up in Audacity, and subtracted the 16-bit clip from the 24-bit clip to get the difference. This represents the actual loss.
You aren’t supposed to hear anything.
Here is that same clip, except amplified by 48 decibels — that same rainfall-to-practically-silent conversion I mentioned earlier. As another way to look at it, if you play the 16-bit song as loud as a rock concert (~115 decibels according to my earlier link), then this is what you’d be missing.
Lost signal (amplified by 48 decibels)
Aside from the fact that it’s still pretty quiet, I also make the observation that it isn’t particularly interesting — it doesn’t match up to the song.
In conclusion, I think this is a crappy marketing stunt by Apple and nothing more. In the best case, people will end up with 50% larger files that sound no different, and in the worst case people might pay to replace their old 16-bit songs, essentially paying twice for what amounts to the same thing. On top of all that, 24-bit files aren’t supported by many devices. Apple would do far better to start selling higher bitrate (or lossless!) files, rather than 24-bit files.