Audio output with USB soundcard 24-bit DAC : who convert audio stream 32 to 24 bit ?

[krisnotoff]

Hello,

Ableton internal processing depth is 32- bit.
But my audio output is with a 24-bit only USB soundcard DAC and an ASIO driver.

Do you know what happens then ?
Does Ableton proceed to an audio treatment (32 bit to 24 bit) before to send the 32 bit digital audio stream to ASIO driver ?
I see nothing about this in the documentation (Please i don't speak of audio recording, ADC is another subject).

[[jur]]

32 bit is used for processing (signals through devices/plugins), which is different from the DAC conversion. It's two different processes.
In short, you shouldn't worry about this.

[krisnotoff]

Thanks, i'm not worry but 32-bit to 24-bit operation is not neutral.
Ableton can't just send the 32-bit internal audio stream to the soundcard if the bit depth is not the same. A 24-bit DAC only accept 24 bit.
Does Ableton use a dither algorythm to go down to 24 bit before the soundcard driver or Ableton ignore that difference of depth and there is just a truncation of the digital audio stream resolution to 24 bit by the driver ?

[[jur]]

Let's think about this:
when you're playing a 32 bit rendered audio file from the Finder/Explorer to your DAC, what happens?

I think the question is not about what Live is doing, it's processing at 32 bit and that's what is being output during playback. You make a decision to downsample only when exporting. (Note that I could be wrong here! Maybe you're right in that Live is doing the reduction and send it to the DAC)
The question is what your audio interface is doing with 32 > 24 bit conversion... and frankly I don't know how it's handled by DAC which afaik are 24 bit for the vast majority. I'd think they apply dither but that's only my undocumented believe here.

What I know though is that I stopped dithering my exports many many years ago because if I can't hear a difference, then it's not worth trying to hear if it really makes a difference (and trying to convince yourself you could probably hear anything) And if you, as a years-long trained mixing engineer/artist can't hear it, then no single music listener will ever hear it.
Some of those files I didn't dither ended up on TV and in theatre, and nobody ever asked me: "I can hear something wrong, are you sure you didn't forget dithering?".
24 bit resolution is a ginormous dynamic range, way larger than the human ears' range. Imo 32bit makes sense for recording audio in some situations, it makes life simpler and safe with very dynamic sources (but then what dynamic range your mic can handle?), and if you want to process audio like crazy and experiment with finding hidden artifacts down the silence range created by squashing it with 20 OTTs.
But as for outputting audio, it's not really relevant imo.

I know it probably does not answer to your question exactly, but yes maybe try asking Support OR your soundcard manufacturer about this, that'd be interesting.