32bit vs 24bit

[HorusProject]

Now then... I have been reading a book on Ableton Live that says its best to work in 32bit and do all your recordings in 32bit because Live's internal processes are 32bit. Hmm i am not sure if this is right, and I am not sure what difference this makes anyway?

I understand working in 24bit leaves more headroom for processing the project when its ready for mastering ok but is 32bit all that good? Is it abit excessive? I mean if all our files where in 32bit depth and 96k sample rate we would be putting much more strain on our resources and lose out on the benefit of clarity in our sounds because of drop outs.

please correct me if I am not making sense.

So... 32bit or 24bit whats are your opinions?

Horus

http://soundcloud.com/horusproject

[n8tron]

As far as I know there are no interfaces capable of recording 32bit audio... I don't believe Ableton, or any other software can record 32bit audio.

[SKerkewitz]

I see no reason why someone should work in 24bit instead of 32bit except that the wave file on the disk is a little bit smaller. From a computer science view 24bit is a bit depth that is not that great to process instead of bit depth like 8,16,32,64 or 128bit.

[HorusProject]

As far as I know there are no interfaces capable of recording 32bit audio... I don't believe Ableton, or any other software can record 32bit audio.

I have used 32bit in the Record bit depth option in the preferences, and the samples I have recorded using a Focusrite Saffire audio device are showing as 32bit in the sample display of the clip view... so it is possible but I dont know if it is any better than 24bit.

The samples I get from Computer Music magazine like last months Vengence-sound are 16bit and sometimes they give out 24bit.

[crumhorn]

I use 24 bit because it allows me to leave a generous gain margin when recording instruments and vocals. I always export the final mix as 16 bit.

Whatever wave file format you use live will convert it to floating point format for internal processing. AFAIK it uses 32 bit floating point in some places and 64 bit in others. it tells you in the Audio Fact Sheet at the back of the manual.

[Tone Deft]

the samples I have recorded using a Focusrite Saffire audio device are showing as 32bit in the sample display of the clip view... so it is possible but I dont know if it is any better than 24bit.

it is not recording at 32 bit, rtfm, I just did.

AFAIK no sound cards at 32 bit and all commercially available audio A/D converter chips are 24 bit.

32 bit talk is for ITB mixing and mastering stuff.

if you have to ask about the difference between the two, it doesn't matter.

[HorusProject]

Saffire Pro 24 DSP has "Supported Sample Rates: 44.1kHz, 48kHz, 88.2kHz, 96kHz"

ok so this is sample rate for recording but it does not say anything about bit depth... if you look at the preferences for Live under recording you get to choose bit depth of 16bit 24bit or 32bit

so why does Live say it is recording in 32-bit and manual does say it has 32-bit architecture?

anyone know more about this?

"To ensure neutral recordings of plug-in instruments and any audio signals that are being processed by effects plug-ins, internal recording at 32 bits is recommended." from Audio Fact sheet in the Live manual

[crumhorn]

That Record Bit Depth option describes the type of audio files that Live will create. It is not necessarily the same as the bit depth delivered by your sound card. If you record from a 24 bit source into a 32 bit file the least significant 8 bits of each sample point will all be zero.

[Tarekith]

Whoa, everyone slow down Simple answer:

Live does a lot of operations like freezing as 32bit files (I believe 32bit float, but maybe fixed point), because they can't be sure that you're not clipping a track when you do the freeze operation for instance. It's my understanding that a 32bit floating point file uses the same 24bits to capture the audio signal that a 24bit file does, the 8 extra bits are there to just capture any signal that would normally exceed 0dBFS in a fixed point file.

Long story short, as long as you're not clipping anything anywhere (and you shouldn't be anyway tsk tsk), there's nothing to be gained from using 32bit versus 24. It doesn't offer better sound quality, only larger file size.

In the Audio Fact Sheet they recommend using 32bit files to remain nuetral, because if you were freezing a track that was clipping, converting it to 24bit would result in a clipped audio file, where as the 32bit floating point file could still capture the signal without clipping.

So yeah, there's nothing wrong with using 32bit files for everything, but if your signal never clipped anywhere, then you're really only increasing for file size by 1/3rd for nothing. I always recommend people use 24bit, and just follow good practices when it comes keeping your levels out of the red.

This is a very simple explanation, if any of the math or dsp wiz's want to correct my semantics, or expand on this, please feel free.

[Tone Deft]

yep.

if you're still in doubt, try the various settings yerself. there are other things that are more important to watch over like not clipping in effects, gain stages etc.

it's a common question, nobody is born knowing this stuff.

[Tarekith]

I think it's a common theme in a lot of digital audio discussion, most people assume "more" of something is is better. While it's rarely worse (we can discuss 192kHz sample rates another day), most of the time it doesn't really add much benefit and only a performance hit. 99.9999% of the world can make absolutely stunning music with just 24bit. 44,100kHz for all their work.

[invol]

I think it's a common theme in a lot of digital audio discussion, most people assume "more" of something is is better. While it's rarely worse (we can discuss 192kHz sample rates another day), most of the time it doesn't really add much benefit and only a performance hit. 99.9999% of the world can make absolutely stunning music with just 24bit. 44,100kHz for all their work.

Exactly.

[HorusProject]

thank you everybody for all your input.

This forum is a great training ground.

[keefbaker]

32bit is pointless. 24bit is perfectly good enough for the dynamic range of the human ear. 32bit means if you can hear the lowest bit of dynamic range then the top will utterly destroy your ears.

Using 32bit you'll be no percievable improvement in quality and all the extra processing will slow your DAW down.

[ian_halsall]

It's the internal resolution which if you do a million computations on 24 bit floating point numbers all the roundings will give a different answer than if you use 32 bits or higher internal resolution.

Imagine rounding everything to 1 dp and then multiply 10 fractions together - different from no rounding clearly.