24bit vs. 16bit Sample Library Comparison

[KenNickels]

Vendors charge more for 24 bit. Why choose 24 bits? I mean, I know in a sort of general theory, but I would like to hear from the community about which aspects of the sound are affected, from a subjective point of view.

Thanks

[twistedtom]

24 bit has better resolution of volume that is all. The sample rate is frequency.

[Kubi]

Just to clarify: While the bit depth is indeed related to the resolution of the amplitude of a sampled wave over time, it is a bit too simplified to say it is about "volume". (And while sample rate is the frequency with which samples are taken, it has actually little to do with the frequency of the material, apart from having to filter all frequencies above half the sampling rate in the source material.)

You can hear the difference very easily yourself:

Take a high-quality 24bit recording, import it into a 24bit DP project. Export it using DP's export function as a 16bit file, then listen to the two side-by-side.

You'll hear a far smoother extended high end, cleaner transients, more and smoother low end, much better imaging and much better depth in the 24bit recording vs. the 16bit recording.

(Be sure that you have "dither" checked before you export as 16bits, otherwise you'll get a truncated 16bit file, and that just sounds plain nasty.)

With sample libraries the difference very much depends on the kind of library and how the samples were recorded in the first place, but generally you can very much hear the same difference you'll hear if you do the test described above.

So while it sometimes is perfectly fine to use 16bit samples (i.e. because you need to fit more samples - 16bit will use 50% less memory than the same amount of 24bit samples, or because you are using a library that was recorded in 16bits to begin with, i.e. older libs, or because you actually like the slight 'crunch' and ever-so-slight lack of high-end and low-end extension on that particular type of sound) it's generally much better to stick with 24bit samples, just like it's much better to stick with 24bit audio recording vs. 16bit.

[twistedtom]

That is true and that all has to do with better resolution of volume, I said it kind of simple as I had no time and meant to add to it but I got busy and forgot.

[KenNickels]

Thanks for your observations. I know that the more bits in the sample, the less aliasing distortion in the conversion. Maybe that is adding to the fidelity you describe.

So, do you usually record in DP at 96kh/24bits and dither down to 44.1kh/16bits? This is better than staying 16 bits the whole time?

[Kubi]

I personally use 44.1kHz or 48kHz SR (for records resp. film/video) but yes, everything in 24bits, always. I then dither down to 16bits after every other step, incl. mastering.

If I don't use an outside mastering house I usually dither down using one of Peak Pro 7's algorithms at the highest quality setting.

If and when I need SR conversion, I do that in Peak as well. That, too, would happen before I dither down to 16 bits. So going from 24 to 16bits is always the very last step, with no other audio processing to follow.

[KenNickels]

I personally use 44.1kHz or 48kHz SR (for records resp. film/video) but yes, everything in 24bits, always. I then dither down to 16bits after every other step, incl. mastering.

If I don't use an outside mastering house I usually dither down using one of Peak Pro 7's algorithms at the highest quality setting.

If and when I need SR conversion, I do that in Peak as well. That, too, would happen before I dither down to 16 bits. So going from 24 to 16bits is always the very last step, with no other audio processing to follow.

Might be a dumb question, but if you record at 24bits in DP, does that require your VI's to all be 24bit as well? Conversely, if you record in 16bits, do the VI's have to be 16bit? Must they match?

[bdr]

Simplified, 24 bit is a much more accurate way of describing each sampled point in the waveform...more numbers get used to describe the 'volume' at each point of the waveform...eg 3.437 is more accurate than 3.5.

[Kubi]

Might be a dumb question, but if you record at 24bits in DP, does that require your VI's to all be 24bit as well? Conversely, if you record in 16bits, do the VI's have to be 16bit? Must they match?

No they don't have to match. Any 16bit sample or instrument inside a 24bit project will start out just using 16bits and leaving the remaining 8 at zero. (There's actually a plug-in that will show you how many bits you are using, have to check what it's called.)

BUT: As soon as you do anything to that audio - change level, EQ, compress, or any other process - it will actually use the full 24bits resolution for the results of the process. So very quickly the 16bit VI or audio track will become an all-24bit-affair inside a given project.

In other words, even if all audio tracks and VIs were 16bit audio to begin with, it's still well worth it to run the project in 24 bits, since chances are you're going to do something to that audio during the work process , and thus you will benefit from the higher resolution.

(Actually, to be precise: the processing will happen in 32bit float resolution, which is essentially 24bit resolution with an added 8 bits for virtually infinite dynamic range. But you'll then write the result of that 32bit float process in 24bit files. If your project is 16bits, the result gets written in 16bit files, losing a lot of resolution in the process. Now, you can set your project to remain in 32bit throughout - costs 33% more storage space and some extra processing power compared to 24bits, but should theoretically yield some gains re. fidelity. For me, I have found using 24bits to be virtually indistinguishable from 32bit float. I assume this has to do with the fact that the extra 8 bits are spent entirely on dynamic range - important for processing but not so much for writing files? Not sure.)

[KenNickels]

I personally use 44.1kHz or 48kHz SR (for records resp. film/video) but yes, everything in 24bits, always. I then dither down to 16bits after every other step, incl. mastering.

If I don't use an outside mastering house I usually dither down using one of Peak Pro 7's algorithms at the highest quality setting.

If and when I need SR conversion, I do that in Peak as well. That, too, would happen before I dither down to 16 bits. So going from 24 to 16bits is always the very last step, with no other audio processing to follow.

You could do all this in DP, right? I'm guessing Peak does it better?

[mhschmieder]

Peak does it MUCH better than DP, but iZotope goes a few steps further still (and they have some revised algorithms that will be part of the upcoming iZotope PRO 5 release).

Of course, it can be program-dependent (where "program" means "material"). Some like one dithering algorithm for classical and another for hard rock, etc.

I do all of my SRC off-line now in RX2 Advanced as a Standalone app. It allows me also to take care of stuff like that while a DP project is still open.

I'm not going to take the time to hunt it down right now, but there are on-line comparisons of SRC algorithms (and dithering as well), that are quite useful and telling, as they show the results at several different settings. Of course, your ears are ultimately the best judge.

Anyway, those links have been listed in this forum a few times as well, so that might be a quicker path to re-finding them than a Google hunt.