24 bit vs 16 bit for midi samples

Question

I haven't been able to find much of a satisfactory answer anywhere regarding the great debate between 24 bit and 16 bit sound recordings. It seems to be that most people say recording in 24 bits is the way to go, because of a larger dynamic range, but I'm wondering if that really translates to better quality in a practical sense that most people will notice.

At any rate, I wanted to know whether there would be a noticeable difference in 24 bit midi samples as opposed to 16 bit midi samples. The midi orchestral plug ins that I use come in 16 bit and 24 bit. I'm wondering if it's worth shelling out the extra money to upgrade to the 24 bit and if I will really notice a huge difference in quality.

I'm using a UAD Apollo as my audio interface. An uno midi to usb cable to connect my keyboard, and I compose orchestral works that range from 4-16 individual midi tracks.

Will 24 bit midi samples and 24 bit recording create a perceptible difference, and is there anywhere that I can check for examples?

Answer 1

In my opinion, the audible difference between 16 and 24 bit audio is minimal at best. I'll do some bullet points because i'm tired:

• 24 bit does enable the file to carry more data, and therefore have a lower noise floor, but the noise floor of 16 bit audio is approximately the same (or lower) as that of most amps and preamps. It's definitely lower than the average listening environment.

• 24 bit recordings may be desirable if you're going to be doing some serious manipulation to the audio. I don't mean dynamics or eq; more like time expansion or pitch shifting. However, there may not even be an audible difference in this case. I've never checked.

• Any non-audiophile audience (who knows, maybe even audiophiles) will not notice a difference between 16 and 24 bit recordings. It's nice to have, and it makes us feel like our files are of higher quality but, pragmatically, the difference is negligible.

• I just ran a test on a voice track because i'm always paranoid about being wrong. It was recorded at 24 bit. I exported it as 16 bit, reimported it, flipped its polarity, and lined it up against the 24 bit original. It actually came in with a -80dBFS noise floor, so i learned something today. This still shouldn't be perceptible unless you're doing some intense processing.

So, to answer your question, 24 bit samples are nice to have, but they're not strictly necessary. 16 bit samples should be totally fine for most purposes. Hope that helped!