32-bit float vs 24-bit for Recording.

Are there any advantage to setting Reaper (or any DAW) to 32-bit float for tracking? Does 32-bit float help prevent overloading and distortion during recording process?

Aside from a 50% increase in file size, are there any downsides to 32-bit float as opposed to 24?

 

With 32-bit float, you might not be able to clip Reaper, but you can still clip your audio interface.

 

32-bit float is just a pure joke (and i already said it several times on this forum ...) for a SIMPLE reason:
there is NO 32bits float AD converters.

ALL converters are 24bits integer

Basically, you are just "repacking" your 24bits integer into 32bits float.
Loosing space and CPU ressources

 

The TRUTH about 32 BIT FLOAT for audio recording– Really useful or marketing hype?

 

Your audio interface is 24bit (some are not even 24bit, more like 20) and it clips (both in and out) at 0dB, as it has 144dB dynamic range.
Internal processing is 32bitFP (unless you set it to 64bit), which doesn't clip as it has 1528dB dynamic range. I think it clips at +728dB or something crazy like that.

 

For recording: All else being equal and assuming you're not clipping on the way in, you can't improve on 24 bit unless every device in your recording chain has a noise floor of less than -144 dBFS - so not realistic.

For mixing: Reaper mixes in 64 bit unless you changed this.

For printing: 32 bit for archiving, 24 for film, many streaming platforms, possibly archiving, 16 for CD.

Since you're on Reaper, you have 32 bit FP WavPack available for your own storage/archiving needs (bouncing, freezing, stems). The size difference between 24 and 32 isn't very dramatic with compression.

 

For audio processing there is no reason not to use it since you convert down to 16 bit later anyways.

People always search for reasons to use the highest sample rate, bit rate etc.. For the technical processing foundations I rather can advice to search for fundamental reasons not to use it. And if those reasons do not hinder - just use the highest option possible.

A bit of simplicity is advisable instead of confusion where it really is not necessary at all.

 

24 bit recording, or 32 bit for modern ADCs, but don't know benefits, advantages...
As a standard, processing from plugins goes either with 32 bit float, or 64 bit float, same with reaper mixing etc.... Floating point is fine virtually, don't know about adc and dac a lot. Go 32 bit float or 64 bit float. I use always 64 bit float, because of plugin dsp quality maximum. Several plugins use exotic 80 bit extended, 117 bit etc, but it is beyond useless fantastic sci-fi math..... Even if reaper adds 80 bit mixing paths and dsp, I will not go with from 64f...not for my cpu.....
In any case, if you release mp3 only, no any lossless hires, and mp3 normalised by loudness lufs standard, you don't need dither... Because it will be killed by lossy mp3 (mp3 is a lossy processing, and dither don't like any (destructive) processing after that) and gained-down volumes (quantization error will appear). So, dithering for lossy and normalizations is just noise, no safety for digital anymore.

 

Also, do we need -1dBTP output, if it will be coded into lossy after normalization to lufs? In any case there will be headroom for peaks..

 

There is a difference between 32bit float inside your DAW or interface. If the interface is NOT 32bit float and the DAW is (or anything higher for processing), if you send clipped audio to the DAW, it will still distort. In this case, the DAW will not clip during playback or monitoring if the clipping occurs inside the DAW. Now, if the interface is also 32bit float, you can send a really hot signal and it will not clip the recording, but the DAW must be set at 32bit float (or anything higher for processing) so that you can recover the audio back. It is worth noting that your audio WILL clip when bouncing, unless you bounce to 32bit float which is useless for anything other than archiving.

 

Last edited: Oct 18, 2024

Reasons to NOT use it :
more CPU
more space
Useless repacking ressources
...
And probably more

The same happened in photography domain :
Fuji do the same so called" marketing hype" ... without any real upside AND more downsides (space, ressources ...)

Some reading :
https://www.dpreview.com/forums/thread/4560530

Sidenote : most modern DAWs audio bus are already 32bits float, even 64bits float ... so imagine the dynamic range :D
https://gearspace.com/board/music-computers/1338544-internal-bit-depth-different-daws.html

 

Last edited: Oct 19, 2024