Sample Library with different bit depth and sample rates

[Stromkraft]

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

Live uses

"double precision (64-bit) summing at all points where signals are mixed, including Clip and return track inputs, the Master track and Racks"

Section "32.2.4 Summing at Single Mix Points" of the Live 9 manual

Does that mean the full track potentially uses 64 bits inside the track strip? It wouldn't appear so. Anyway, what was measured by Bitter here was bits used, so hard to tell. The important thing to me here was to determine whether I was correct saying that a lower bit source is upsampled in Live when you process it.

[Stromkraft]

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

Short answer then is that I don't worry about this.

When I process the audio from my drum rack the source bit depth are already at 32bit floating point going into the effect plug-in or getting to be mixed (in 64bit at the mixing point). In addition the sample rate in the audio stream is according to the rate settings in the preferences of Live and not according to the source material.

I found your questions interesting enough to warrant some investigation on the audio stream in Live, as you see in previous recent posts. I feel even more confident now this is a no problem. As 96000 sample rate samples naturally are bigger in size and, as I assume Live need to throw away information when using any 96000 recording in a project with lower sample rate and because of this they might be unwieldy to handle effectively. But that's hardly a reason to start a big converting operation. Build your library instead,

[Angstrom]

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

Live uses

"double precision (64-bit) summing at all points where signals are mixed, including Clip and return track inputs, the Master track and Racks"

Section "32.2.4 Summing at Single Mix Points" of the Live 9 manual

Does that mean the full track potentially uses 64 bits inside the track strip? It wouldn't appear so.

it uses 32 bits inside the strip (or stream), and then when those are mixed together such as when the channel faders all route to the master, that's a "mix point" and it uses 64 bit for that.
32 bits is quite a large dynamic range (about 96db) and most people would be surprised how low in volume a 32 bit signal can get and still be resurrected.
It's worth pointing out that this isn't an absolute range "below 0db" but a floating window of possibility so if the loudest signal in the 32 bit floating point stream hits -18db then the dynamic range will stretch from -18db all the way down to -114db !!

So 32 bit is fine for a single stream of processing. When we start to mix different "absolute" ranges together, so one track peaks at -18db and another is at -3 we may require the capabilities 64 bits. Hence "at all mixing points" , or whatever phrase it is they use.

I hope that made sense. I'm not qualified to talk about these matters, just stuff I picked up and perhaps incorrectly.

[greyfloor]

TAREKITH: You have a good point. I may say after reading it, I decided from here that the best for me it will be work with 24bits/44100Hz. Thank you for your explanation, it helped to see the thing more clear.

STROMKRAFT: That's what a call an invesgative spirit pal! Considering what you've researched and what all you guys have mentioned, it's not necessary to be worried regarding a sample library with different bit depths and sample rates, at least if working with the last version of Live… it seems the daw is more than competent working with different resolutions materials in real time and downsampling. But my last doubt is regarding upsampling. Working in a 24bits/44100Hz project, with my audio interface ajusted with the same setting preferences, when using the 16 bits/44100Hz samples, I will not have any technical problems at all, is that right? It's better to use these samples trought drumracks/sampler instead using them as a "pure source" as they are, like inside a clip on session view/track??

By the way, thank you so much for all this debate. It's quite nice not only because my particular doubt but mainly because this specific question is hardly found in other topics and foruns.

[greyfloor]

Nobody with some knowledge about this matter to help me with this last doubt??

[Stromkraft]

But my last doubt is regarding upsampling. Working in a 24bits/44100Hz project, with my audio interface ajusted with the same setting preferences, when using the 16 bits/44100Hz samples, I will not have any technical problems at all, is that right? It's better to use these samples trought drumracks/sampler instead using them as a "pure source" as they are, like inside a clip on session view/track??

I think you will have no technical problems whatsoever. I think it wouldn't matter where you are using these samples, but a sample may not sound the same in Sampler/Simpler as the defaults are different.

[fishmonkey]

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.

according to the manual the SOX algorithm is only used when rendering, i.e. for offline downsampling.

32 bits is quite a large dynamic range (about 96db) and most people would be surprised how low in volume a 32 bit signal can get and still be resurrected.
It's worth pointing out that this isn't an absolute range "below 0db" but a floating window of possibility so if the loudest signal in the 32 bit floating point stream hits -18db then the dynamic range will stretch from -18db all the way down to -114db !!

actually 16 bits gets you about 96 dB of dynamic range, 24 bits gets you about 144 dB, and 32 bits is about 193 dB...

[Stromkraft]

actually 16 bits gets you about 96 dB of dynamic range, 24 bits gets you about 144 dB, and 32 bits is about 193 dB...

There's a more detailed overview on Dynamic Range and Signal-to-Noise Ratio in the book Digital Audio Essentials by Bruce and Marty Fries.