Ableton exporting

[stone123]

Hi

I know this question has been asked before, but I haven't found a definite answer yet, so my apologies.

I've been pondering lately about what's the best settings for project and exporting. I'm on Windows 8.1 and Ableton 9.5. My external audio card is Lexicon's Alpha and is restricted to 24bit bit depth and max 48 kHz sample rate.

My project settings (concerning bit depth and sample rate) are: record mode 24bit, sample rate 48 kHz. Audio device is Alpha ASIO (which, as stated, only allows max 48 Khz sample rate for project). I'm not recording anything externally (using mics or audio input).

Mastering will be done in a different process.

My questions are:

1. Is it any good to export at 32bit bit depth?
2. Is it useful to export at 96 kHz sample rate?

So what's best: 24/48, 32/48, 24/96, 32/96?

I've successfully exported at all the rates, and they all sound equally good or bad. I don't see why upsampling to 96 kHz would be any better than 48 kHz.

But if I had a better soundcard, that could do 24bit/96kHz, would it change anything? Could I use higher project settings, would the exported audio qaulity be better? Does it affect the situation at all?

stone

[RichieWitch]

As I understand it, exporting at a higher resolution than what your A/D converters use will actually degrade your sound. The reason for this is that you are starting with digital bits and sample rates and bit depth. When you increase any of these computationally, the algorithms must fill in the tiny blank spaces you are creating in the data. And since it's not part of your original signal, what it fills those spaces with can only be considered "noise".

Now... with that being said, Ableton uses 32-bit floating point bit depth for all its processing, regardless of your settings, so you could experiment with that and see how it sounds.

For a really detailed explanation of how all this works, check out "Digital Audio Explained For the Audio Engineer" by Nika Aldrich.

http://www.amazon.com/Digital-Audio-Exp ... 141960001X

[stone123]

So, you suggest that exporting at 32/48k would be best in Ableton? And with better A/D converters (soundcard?) higher rates would be better?

[Tarekith]

If you're not clipping the master channel at all (and why would you be?), then 24/48 is fine. Ableton's upsampling algorithm (downsampling uses SOX) is only so so, bordering on not good, so I wouldn't bother with 96. 32bit won't hurt, but probably isn't necessary either.

[stone123]

Thanks Tarekith, -6 dB or so and been working with those safe settings (24/48) so far. But if I bought a better soundcard, would I benefit any in the quality? Mastering will be done with Ozone (5/6/7) or mastering house. Does that change anything?

[RichieWitch]

So, you suggest that exporting at 32/48k would be best in Ableton? And with better A/D converters (soundcard?) higher rates would be better?

No, I'm saying you can go downward (e.g. from 24 bit to 16 bit, or 48k to 41.1k sample rate) without any unusual issues, but when you start moving in the other direction, raising sample rates and bit depths during exporting, the D/A converters have to insert bits that did not exist in the original material.

It's easier to hide the removal of bits (via dithering, etc), but adding bits is another matter.

Consider too that, mathematically speaking, you can only recreate frequencies that are equal to or less than half of the original sample rate. For example, at a 48k sample rate, 24k Hz is the highest frequency that can be reproduced accurately.

So moving from 48k to 96k is adding an enormous amount of extra data without adding any accurate sound to the source, since the extra frequency content that 96k sample rate enables (up to 48k Hz), does not exist in the source your exporting (capped at 24k Hz).

[Tarekith]

Thanks Tarekith, -6 dB or so and been working with those safe settings (24/48) so far. But if I bought a better soundcard, would I benefit any in the quality? Mastering will be done with Ozone (5/6/7) or mastering house. Does that change anything?

Your sound card doesn't matter at all for this kind of stuff, it all happens on the computer itself and never has any reason to involve other peripherals.

[stone123]

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?

Ok, Tarekith, I guess you mean that all the computing are done in the box, all the sample rate and bit depth conversions? And the D/A (A/D) conversions happens in the soundcard? But that's only when you want to hear something? Or input some sound.

Going downwards in bit depths seems to need dithering, but does going upwards hurt (other than that the file size grows)?
And, does it hurt going upwards in sample rates (other than that the file size grows)? And loads of useless calculations?

Obviously going upwards is useless in both respects, or what?

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?

[Tarekith]

They mastering house is asking for 96k? That sounds odd to me, never heard of that myself, especially as they could convert it to 96k better than you could most likely (no offense).

And yes, the soundcard doens't play a role because the only time it's needed it when you convert your digital signal to analog. A better soundcard can of course do this better, and thus you might make mix decisions with more accuracy. But in terms of actually exporting your song, it has no influence on the sound quality.

If you've already decided on a mastering house and they MUST have the files in 96k, drop me a message and I'll convert it for you with a high end SRC.

[stone123]

I guess I must have picked the 96k asking somewhere in the threads, so it's probably not valid assumption.

But what if I had loops, samples, recordings of higher sampling rates than 48k, say 96k? Would it then be better to export at higher sampling rate?

[zethus909]

48k sample rate, 24k Hz is the highest frequency that can be reproduced accurately.

what? i thought sampling rate is how many times per second...an audio source is being sampled....i didnt think this had anything to do with what frequency content you can capture......how does that make sense.....

the human ear can only hear up to 20 khz etc...ok whatever...but

if there is cymbal crash...you will capture all the frequency content of it....the highest and the lowest...no matter what sampling rate you are using..

you could be using 22,050 or 96k and you will capture the highest frequency content of that audio it doesnt make a difference...

the difference is that you would only be capturing that audio 22,050 times per second vs 96,000 times...so you would get a better representation of that audio source...if that audio source is being captured directly through the air through a microphone...at 96kz......then you would technically want to convert that out of ableton at 96 khz so you dont lose any of the actual audio content (not how much of the highs you caught..just how many times you took a sample of it) the less times you take a sample per second of whatever frequencies..the less good reproduction of that audio signal you will have but you wont LOSE any of the HIGH frequency content......right?.......the difference between 96 and 48 etc is almost non existent though i think.....only on the best quality headphones or system cuold you even tell the difference

[Tarekith]

Personally I use 44.1kHz for all my audio work. There are some instances where 96k can sound better if a softsynth is aliasing at lower sample rates for instance, but it's very rare these days I find. In general unless you have a specific and vaid reason to work at higher sample rates, then there's no reason to. Bigger is not always better, though this is one of those things people will argue about for months online.

[stone123]

zethus: That's what I thought too. Even if the final product after mastering will be 16bit/44.1k, wouldn't it be good to keep all the information and export as 96k? Even if the project had also audio material in lower sample rates, say loops in 44k? Would those then be upsampled to 96k?

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?

[whyterabbyt]

48k sample rate, 24k Hz is the highest frequency that can be reproduced accurately.

what? i thought sampling rate is how many times per second...an audio source is being sampled....

Yes, it is.

i didnt think this had anything to do with what frequency content you can capture......how does that make sense.....

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?

the human ear can only hear up to 20 khz etc...ok whatever...but

if there is cymbal crash...you will capture all the frequency content of it....the highest and the lowest...no matter what sampling rate you are using..

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.

you could be using 22,050 or 96k and you will capture the highest frequency content of that audio it doesnt make a difference...

No you wont, and yes it does make a difference.

the difference is that you would only be capturing that audio 22,050 times per second vs 96,000 times...so you would get a better representation of that audio source...

Yeah, 'better' as in 'actually including any frequency content above 10 kilohertz'.

[whyterabbyt]

As I understand it, exporting at a higher resolution than what your A/D converters use will actually degrade your sound.

Erm, no.

The reason for this is that you are starting with digital bits and sample rates and bit depth. When you increase any of these computationally, the algorithms must fill in the tiny blank spaces you are creating in the data. And since it's not part of your original signal, what it fills those spaces with can only be considered "noise".

Reads like your confusing exporting with a specific process (dithering) which deliberately adds noise to improve sound process of exporting to a lower sample rate. Converting to a higher bit depth can be done easily without changing the sound in any way, you're just not getting any actual new information out of those extra bits.
0.5 of a 0-1 scale in 16 bits is 0.5 of a 0-1 scale in 24 bits.