Ableton exporting

[whyterabbyt]

Richie, Tarekith, I'm a bit confused here. If the a mastering house suggests that you send them tracks of 24/96k quality, how would I achieve the 96k sample rate then?

Think of it this way : Your DAW's capture and playback samplerate is determined by your soundcard. But when the DAW exports, it doesnt(*) use the soundcard, so its not limited by it. Thus you can export at any supported sample rate and bit depths, which might be a higher rate than you can actually listen to.

(*) assuming an all in-the-box project

So, I guess, if I don't have sound material that is more accurate than 48k, it's pointless to use higher sample rates in any respect?

Basically, yes, you dont add any new accuracy by upsampling existing material.
However if your project uses synths, then exporting at a higher sample rate should mean that those synths do their calculations at the higher sample rate, which will be more accurate (and include less aliasing etc). That in itself might be worthwhile.

[stone123]

Think of it this way : Your DAW's capture and playback samplerate is determined by your soundcard. But when the DAW exports, it doesnt(*) use the soundcard, so its not limited by it. Thus you can export at any supported sample rate and bit depths, which might be a higher rate than you can actually listen to.

This much I've captured so far!

(*) assuming an all in-the-box project

So, I guess, if I don't have sound material that is more accurate than 48k, it's pointless to use higher sample rates in any respect?

Basically, yes, you dont add any new accuracy by upsampling existing material.
However if your project uses synths, then exporting at a higher sample rate should mean that those synths do their calculations at the higher sample rate, which will be more accurate (and include less aliasing etc). That in itself might be worthwhile.[/quote]

I suppose this supports the idea that if the project contains any higher sample rate material (than 44/48k), then it would be reasonable to keep that information by exporting at higher rate. And do the mastering with all that information, and then finally export the final product at 16/44.1k or whatever the target is.

[Tarekith]

tarekith: so you don't think there is a reason to use higher sample rates in exporting even if the project would contain higher rate material?

The problem like I mentioned is that then Live would do the upsampling to 96kHz during the export, and Live's upsampling is not that good at all. I'd rather downsample with it to 44.1kHz, because then it will use the excellent SOX algorithm and ultimately sound better than the 96kHz export.

If you wanted to work at 96kHz in the future, I would use a third party SRC to change everything to 96Khz ahead of bringing them into Live.

If you have a project that's already got a mix of audio files at different sample rates in it, render at both and listen for yourself too. Go by what sounds good, not the theory.

[whyterabbyt]

I suppose this supports the idea that if the project contains any higher sample rate material (than 44/48k), then it would be reasonable to keep that information by exporting at higher rate. And do the mastering with all that information, and then finally export the final product at 16/44.1k or whatever the target is.

Pretty much.

Technically, by the way, I'd consider that export to your target rate to be the last part of the mastering process, since you'd usually include dithering at that stage. ie the project is 'balanced and mixed' as a whole, and then individual master output files at whatever target rate you need are done from that.

(FWIW, 'mastering' as a term basically pertains to/originates from 'creating the master copy'. When the output was vinyl, the process was originally more to do with dealing with vinyl's restrictions as a medium; only more recently did it become the 'finalisation' of the mix etc that it is today.)

[stone123]

tarekith: so you don't think there is a reason to use higher sample rates in exporting even if the project would contain higher rate material?

The problem like I mentioned is that then Live would do the upsampling to 96kHz during the export, and Live's upsampling is not that good at all. I'd rather downsample with it to 44.1kHz, because then it will use the excellent SOX algorithm and ultimately sound better than the 96kHz export.

If you wanted to work at 96kHz in the future, I would use a third party SRC to change everything to 96Khz ahead of bringing them into Live.

Ah, I see! Ok, now I understand. Thanks! Does Live's upsampling's lower quality apply to the newer versions too?

If you have a project that's already got a mix of audio files at different sample rates in it, render at both and listen for yourself too. Go by what sounds good, not the theory.

No, I don't. All the loops and samples are 44/48k. It was only a hypothetical case.

When is it meaningful then to export at 32bit depth?

[stone123]

Technically, by the way, I'd consider that export to your target rate to be the last part of the mastering process, since you'd usually include dithering at that stage. ie the project is 'balanced and mixed' as a whole, and then individual master output files at whatever target rate you need are done from that.

Yes, that is the idea.

[whyterabbyt]

When is it meaningful then to export at 32bit depth?

As an output format for listening, not so much.
But generally, 32-bit is 32-bit floating point (ie numbers with decimal places). 16-bit and 24-bit are integer (ie whole numbers only). Most DAWs and plugins etc do their calculations in 32-bit (and sometimes 64-bit) floating point maths. So there's already a resampling-for-bitdepth process involved in whatever reaches your 16bit or 24-bit soundcard.


Hence 32-bit float output might be worthwhile if, eg, you wanted to work between more than one bit of software and preserve absolutely the internal audio as it exists 'inside' the DAW without that resampling-for-bitdepth.

[zethus909]

Because you cant properly measure something that happens more frequently than you measure it.

Its science. Proven and understood. You must sample a signal at a sample rate of (slightly over) twice its frequency to be able to be able to reconstruct it.

How does it not make sense?

because the audio signal is a continuous wave?...so when you take a sample...of it....it doesnt matter how often...your essentially taking a scoop out of the flow of it.....now if the wave was not continuous but a digital representation..then you would have to sample at lightly higher than it because you would be missing the start/end of the cycle each time you try? so you would need something with faster cycles to capture that digital info......but this is an audio signal...so thats why what your saying doesnt make sense.....to me....

No you wont. If the highest frequency is happening 100,000 times a second, how the hell do you expect measuring it 10,000 times a second to work out what's happening ten times a sample.

because each time you take a sample...your taking more than one cycle.....youre taking middle of cycles you can take any part of the continuous signals...?

[Tarekith]

Does Live's upsampling's lower quality apply to the newer versions too?

When is it meaningful then to export at 32bit depth?

Yes, currently Live uses the same lower quality SRC for upsampling that it always has. The SOX downsampling algorithm was added in Live 9.

32bit is useful if you're too lazy to make sure you master channel is not clipping before you export. Technically it's the best format to export in, but I've never heard it sound better than 24bit myself, that's just in theory. The basic premise is that by being floating point, it's almost impossible to clip a 32bit signal. Where as 24bit is fixed point, and if your master channel clips, the file when rendered is permanently clipped too. Leave a bit of headroom and it's a coin toss as to which you use IMO.

Lots of crazy and confusing talks about digital audio in this thread, this video is a good primer if you want an easier explanation of the basics:

[whyterabbyt]

because the audio signal is a continuous wave?...

Which is being measured at discrete time intervals.

so when you take a sample...of it....it doesnt matter how often...your essentially taking a scoop out of the flow of it....

No, you're measuring its amplitude at that point in time. At <samplerate> points in time per second.

.now if the wave was not continuous but a digital representation..then you would have to sample at lightly higher than it because you would be missing the start/end of the cycle each time you try?

I have no idea what that is supposed to mean. You wouldnt be sampling a digital signal, because sampling is the process that turns an analog signal into a digital one.

We're talking about sampling an analog signal. A signal at a given frequency N is not a static value, its a changing waveform that has to reach a minimum value, then cross zero, then reach a maximum value, then return to its minimum, and cycle like that N times a second.

To properly(*) sample an analog signal which includes waves with frequencies up to that specific frequency, you need to be able to differentiate between the lowest point of that wave and zero and the highest point of that wave within one cycle of the wave. That means you need to measure more frequently than the wave occurs to see that cycle.

(*) where properly means you include enough information to reconstruct that signal again, ie its full cycle.

As I said before, this is proven science, its called the Nyquist Theorem, google it. Its not a debate. You need to measure at a sample rate of just over twice the highest frequency you want your digital sample to contain.

http://music.columbia.edu/cmc/MusicAndC ... /02_03.php

so you would need something with faster cycles to capture that digital info......but this is an audio signal...so thats why what your saying doesnt make sense.....to me....

Again, you're making no sense. Yes, its an audio signal, one that you're turning into a digital one by measuring it. You dont 'capture' any digital info whatsoever, your digital info is the measurements of your analog signal. And you dont 'capture' it faster than you're capturing it.

No you wont. If the highest frequency is happening 100,000 times a second, how the hell do you expect measuring it 10,000 times a second to work out what's happening ten times a sample.

because each time you take a sample...your taking more than one cycle.....youre taking middle of cycles you can take any part of the continuous signals...?[/quote]

No, each sample measures once. Thats what a sample is. A single measurement only measures what the signal is at that point in time. If the signal changes ten times before another measurement occurs, you have lost those changes, and you cannot even determine if they occurred.

You're trying to say you can measure millimetres with a ruler that only has a metre scale on it. It doesnt work like that.

[stone123]

So to sum up things learned so far:

Workflow: Export mixed project from DAW (Ableton) to be further processed in a separate mastering project.

1. Soundcard, and it's quality, doesn't play a role in exporting a project. Though a better soundcard might help to hear things clearer and therefore help to make better mix judgements.
2. If the project doesn't contain audio material of any higher than 44 or 48kHz sampling rate, then there is no need to export at any higher rate than 44 or 48kHz.
3. If the project contains material of higher sampling rates, then it might be reasonable to export the project in a higher rate, eg. 96kHz.
4. But if the project contains mixed material (some audio in 44k and some in 96k, for example), then due to Ableton's not-so-good upsampling algorithms, it might not be reasonable to export at higher than 44 or 48k, unless one upsamples all the audio to higher rates using third party SRC before bringing them in to the project. (tarekith)
5. I need to learn to quote people correctly!

Conclusion: I keep on exporting mixed projects at 24bit/48kHz.

[whyterabbyt]

So to sum up things learned so far:

Workflow: Export mixed project from DAW (Ableton) to be further processed in mastering project.

1. Soundcard, and it's quality, doesn't play a role in exporting a project.
2. If the project doesn't contain audio material of any higher than 44 or 48kHz sampling rate, then there is no need to export at any higher rate than 44 or 48kHz.
3. If the project contains material of higher sampling rates, then it might be reasonable to export the project in a higher rate, eg. 96kHz.
4. But if the project contains mixed material (some audio in 44k and some in 96k, for example), then due to Ableton's not-so-good upsampling algorithms, it might not be reasonable to export at higher than 44 or 48k, unless one upsamples all the audio to higher rates using third party SRC before bringing them in to the project. (tarekith)
5. I need to learn to quote people correctly!

Conclusion: I keep on exporting mixed projects at 24bit/48kHz.

sounds about right.

[Tarekith]

Perfect.

[stone123]

But generally, 32-bit is 32-bit floating point (ie numbers with decimal places). 16-bit and 24-bit are integer (ie whole numbers only). Most DAWs and plugins etc do their calculations in 32-bit (and sometimes 64-bit) floating point maths. So there's already a resampling-for-bitdepth process involved in whatever reaches your 16bit or 24-bit soundcard.

Hence 32-bit float output might be worthwhile if, eg, you wanted to work between more than one bit of software and preserve absolutely the internal audio as it exists 'inside' the DAW without that resampling-for-bitdepth.

So, would it be meaningful to preserve the 32bit format to the mastering process (T-Racks, Ozone, mastering house)?

[whyterabbyt]

But generally, 32-bit is 32-bit floating point (ie numbers with decimal places). 16-bit and 24-bit are integer (ie whole numbers only). Most DAWs and plugins etc do their calculations in 32-bit (and sometimes 64-bit) floating point maths. So there's already a resampling-for-bitdepth process involved in whatever reaches your 16bit or 24-bit soundcard.

Hence 32-bit float output might be worthwhile if, eg, you wanted to work between more than one bit of software and preserve absolutely the internal audio as it exists 'inside' the DAW without that resampling-for-bitdepth.

So, would it be meaningful to preserve the 32bit format to the mastering process (T-Racks, Ozone, mastering house)?

Personally, I think the difference is likely to be below any sane threshold. If you were absolutely determined to reduce any possible intermediate stages (or utterly anal/one of those audiophile nutters), then go for it, but I'd consider it moot.