Ableton Forum

Hey Guys

I've reading in different forums, there are so many opinions and at the end I'm still confused... If someone could give some clear help about this subject, it would be great, really.

I have a big library with samples, many of them are 16bit/44.100, many others 24bit/48000 and some, 24bit/96000.
I always worked my projects on 16/44.100 but now I want to work on 24bit/48000. As far as I could get, the ideal is to convert all my samples to this bit depth and sample rate. My doubts are:

1) Can I convert all my 16bit/44.100 samples to 24bit/48000 with a good specific softwares without any problems (like artifacts, noise, etc)? I know that action will not increase the quality of these samples, but will make them equal to the other samples I have and besides, people mentioned that working in 24bit, gives more room to mastering because the audio dynamic range. Is that so?

2) When I convert my 24bit/96000 samples to 24bit/48000 I'm going to have to apply some dithering. My projects will be 24bit/48000. Now, if I need one day to create a CD, I will need to resample once more but this time to 16bit/44100 (with some dithering again). Isn't that bad, I mean, that action will not not bring me artifacts to the final material??

3)Trying to work with 24bit/44.100 instead 24bit/48000, wouldn't be a better choice? Is there any big difference on quality besides size?

4) Is there any relevant configuration on Ableton preferences panel which I need to pay attention? I have also to change my M-Audio Audio Interface configuration to the same specifications?

nobody?

greyfloor wrote: 1) Can I convert all my 16bit/44.100 samples to 24bit/48000 with a good specific softwares without any problems (like artifacts, noise, etc)?

That depends on how you do it and also on the purpose of upsampling, but in theory yes. But you should also note that

Live‘s internal signal processing is all 32-bit, so applying even a single gain change makes the resulting audio 32-bit as well — even if the original audio is 16- or 24-bit.

33.3.4 Dithering (Live Manual).

Sample rate conversion is always non-neutral. When up-sampling you shouldn't use dithering and when down-sampling you should choose the dithering algorithm best for the job and ideally only dither once. Read member Tarekith's article New Dither Examples for perspectives on some of this.

You'll find a lot more info in the Audio fact sheet of the manual on the whole thing about audio in Live.

greyfloor wrote:
2) When I convert my 24bit/96000 samples to 24bit/48000 … will not not bring me artifacts to the final material??

Of course it will. You might be able to live with that degradation though even if it's there.

greyfloor wrote: 3)Trying to work with 24bit/44.100 instead 24bit/48000, wouldn't be a better choice? Is there any big difference on quality besides size?

It's not big at all I'd say, but that's really program dependent. I'd also say the difference in resulting size is too small to worry about. A valid reason to choose 48Khz is if this is preferred in the audio interface.

greyfloor wrote: 4) Is there any relevant configuration on Ableton preferences panel which I need to pay attention? I have also to change my M-Audio Audio Interface configuration to the same specifications?

From the manual:

follow the program’s [Live] built-in “Setting up Audio I/O“ lesson

I'd make sure to only use as many inputs and outputs that you are using. You should strive to use the same basic settings, if you can affect this in detail in your audio interface. But I'm not familiar these days with M-Audio.

All of your questions makes me think I should try 32bits in my projects again. Using only 24bit/44.1Khz in my source files now. My last machine couldn't take 32bit projects.

Thank you Strom for this nice explanation.

But I have two more doubts if you don't mind… You mentioned about working with 32bit, instead 24bit. What are the main advantages about this choice? I mean, it's possible to have more room to work at mixing and mastering, but in the other hand, that wouldn't be extremely high CPU consuming and the final size of the audio files wouldn't be huge? In this case, it would be also necessary to convert all your samples (16 or 24) to 32 bit, right? Besides at the final process, if by any case, you need to downsample the final audio project to 16 bit, this woudn't bring more artefacts??

greyfloor wrote:Thank you Strom for this nice explanation.

But I have two more doubts if you don't mind… You mentioned about working with 32bit, instead 24bit. What are the main advantages about this choice? I mean, it's possible to have more room to work at mixing and mastering, but in the other hand, that wouldn't be extremely high CPU consuming and the final size of the audio files wouldn't be huge? In this case, it would be also necessary to convert all your samples (16 or 24) to 32 bit, right? Besides at the final process, if by any case, you need to downsample the final audio project to 16 bit, this woudn't bring more artefacts??

My main point is that you're most likely working with 32bits in the internal mixer audio stream anyway. At least I believe this is the case no matter if you create a 16bit, 24bit or 32bit project. I haven't studied this aspect in detail.

What I believe we're talking about when we differ between projects is the bit depth of the external as well as internal recordings. These are to the best of my knowledge up-sampled to 32bits floating point in playback, even as the source remain in the original bit depth.
I have not conducted experiments for bit depth so I'm not 100% sure, but reasonably it should be close to what I described.

Strom, thank you for the answers! But I'm still confused about my doubt number 2:

I need to convert all my library at the same bit depth and sample rate and wen I make an offline conversion of my 24bit/96000Hz samples to 24bit/48000Hz, I have to apply some dithering since I'm downsampling. How I do, if I need to end the 24/48000 project in a CD?? In this case, I would have to apply some dithering again, since it's needed to downsampling again… And that's not good at all…

How do you guys deal with your sample library with different resolutions?? I mean, how is the off line process, how do you do??

By the way, dithering is applied only at downsampling but not upsampling right?

And when I need to change only the bit depth (but not the sample rate) it's also necessary dithering?

You should not do any of what you are talking about, because you don't understand it.
When you don't understand it how are you going to make things better?

In fact you have it all backwards

How do you guys deal with your sample library with different resolutions?? I mean, how is the off line process, how do you do??
By the way, dithering is applied only at downsampling but not upsampling right?
And when I need to change only the bit depth (but not the sample rate) it's also necessary dithering?

Dithering ONLY relates to bit depth, correctly called word length. Dithering happens at the point where you truncate a long wordlength to a shorter wordlength.

Do not think about adding more word length to 16 bit files, it wont benefit you in any way at all. If you run a 16 bit file into a 32 bit signal path think of that like a 6 foot man walking in a 32 foot room. The room height is there so the man can jump up and down without banging his head. What benefit is there of putting the 6 foot man on stilts just to make him artificially tall? Sure if you had a 32 foot tall man to start with, then this man would require the full height just to fit him in. But to add stilts to a short man? It does nothing, no processing is helped, no sound quality is improved, nothing.

I could go on.

You ever see those people who get Japanese words tattooed on their back which they don't understand?
Don't be that guy.

Have you heard of cargo cults who wave sticks around hoping to attract cargo planes?
Don't be that guy.

I recommend sticking with Live's internal SR resampling algo which was greatly improved in recent versions. The benefit being : it is non-destructive so any mistakes a person who didn't know what they were doing would not matter. The output sound is perfectly fine, especially considering you don't know what it is.

Conversely - Any benefit you might possibly possibly gain by up-sampling offline in a dedicated app and rebuilding library paths would be undermined by the huge process and potential for disaster, the required informed choice of algo and the lack of knowledge of the true task.

Tl;dr Don't do anything. Live handles this automatically, and although not 100% perfect, it is better in the long term than waving sticks at the sky.

Why do you want to work at 48k instead of 44.1k?

I very much doubt OP can SR convert better than Live

http://src.infinitewave.ca

Note for people who don't understand the graphs. Ableton Live 9.11 (and higher) does the best for Sample rate conversion.

Keep in mind that Live 9 only uses the new improved SOX resampling algorithm when downsampling. If you're upsampling, it still uses the rather meh quality one it always used.

Tarekith wrote:Keep in mind that Live 9 only uses the new improved SOX resampling algorithm when downsampling. If you're upsampling, it still uses the rather meh quality one it always used.

Ah, I had either forgotten or ignored that detail.

Thank you guys for the attention about this doubt! I've started this topic because I've read so many opinions in other foruns, and couldn't find a real and concrete idea about this subject, what makes me a little bit surprise since it's a common situation (having a non homogeneous sample library) and nobody so far could explain a direct method to deal with this situation… I also like to hear different opinions because it's good to have different point of views.

ANGSTRON: Thank you for the help my friend! I only have to desagree with you about one thing… When we don't know exactly what we have to do, we will surelly do things better if we want! Because in my particular case, where I'm in doubt about this subject, I'm here to ask for people who knows more than me, and from this I can learn a lot, don't you agree? Specially if I hear a lot of point of views. If I'm still in doubt or curious I need to ask… There are a lot of opinions about this subject, where many say it's far better mixing in 24bits than 16bit - just because of the dynamics and how vsti has more information to work. Upsampling 16 bit to 24 bit surely will not bring more definition to the sound but will make these equal to the others I'm going to use and the project which I will work. It will avoid a lot of realtime processing or possible conversion problem. As far I could read in other places, people say it's far better convert everything offline but I think we learn a lot listening different opinions. Just to understand your method: the right thing in may case it would be to use all kind of different sample with different resolutions in the same project ( 16bit/44.100Hz, 24bit/48000Hz, 24bit/44100Hz and 24bit/96000Hz inside a 24bit/48000Hz project) and let Live do all the work in real time; then if I need finish apply dithering in the end? This will not bring me any problem (even if I work with more than 100, 200 samples for instance) ? Because my only concern is the technical aspect, not the criativity. Please, my friend don't take me wrong, I'm not trying arguing with you I'm just asking to learn ok !! By the way, I don't use tatoos or like cults : )

TAREKITH: So Tare, as I don't know yet in so many ways I can use the final material which I'm working on, I would like to work on 48000 because perhaps I can use also in a video material (youtube, DVD, etc…). How about your opinion? By the way, how about a different sample library, how would you deal with it to let all samples with the same resolution? Would you convert all of them offline or would you prefer to work in the way Angstron explained to me?

Guys, please tell me how would be your methods to let a different sample library with the same resolution. I would like to know more about it!

In general it's usually a good idea to avoid doing more sample rate conversions that you need, as it can negatively impact your audio quality. It's doubtful you'd be able to hear it once or twice, but the more you do it, the more you run the risk.

The good thing about using Live, is that you can use audio files with multiple sample rates at the same time, and in Live 9 at least, it will use a very good sample rate conversion algorithm to downsample to the project sample rate. So if you keep the same rate in your project set to 44.1, then no matter what your sample library samples are currently, you know they'll be converted with one of the best options out there.

The same is not true if you start working with your project set to a higher sample rate, like 96k. Live will still convert other your samples with lower sample rates to match this, but the algorithm is honestly not the greatest. In this case, it makes sense to me to just stick with using 44.1k while you write your tracks, and let Live do it's thing in the background to make sure all the sample rates match.

You mention MAYBE wanting to work at 48k for video in the future, but unless you know you're going to be doing a lot of that as your main focus, I don't think it's worth converting everything just in case. You can always convert your mixdowns after the fact using a third party conversion tool if you really need to.

Personally, I still work myself and recommend people use 24bit/44.1k for everything unless they have a pressing need to use other SR's. There's a few very rare instances where a higher sample rate might come in handy (preventing bad aliasing, etc), but I don't think the supposed increase in audio quality they MIGHT offer usually survives downsampling back to 44.1 anyway if that's your target goal most of the time.

Stromkraft wrote:you're most likely working with 32bits in the internal mixer audio stream anyway.

As a little follow-up test I recorded 3 files at 44.1Khz in a "Recording" track. One 16bit, one 24bit and one 32bit. I tapped this to a muted bit measure track. I placed the plugin Bitter from Stillwell Audio on each track including the master.

Bitter shows the bits used in the audio and is not a reflection of the capacity of the DAW nor the source file bit depth.
According to the readme coming with Bitter

"the audio host program will reset the bit depth whenever it processes the signal at all" [including] "any nonzero fader setting"

In Live I found that

• fader levels does not change the bit depth in playback, so an unprocessed 16 bit file displays as 16bit regardless of the fader.
• As soon you're processing with a plug-in or audio effect, the amount of used bits go up.
• If you route the 16 bit source track to a new track the amount of used bits in the stream goes up.
• The same in the master.

Here's an illustration with the 16bit source track "Recording", the track "bit measure" receiving audio from the previous as well as the master. I assume that the reason the meter goes above 32bits used is because Live uses 32bit floating point.

Live uses 64 bit at all mix points, and 32 bit in the stream itself.