I, like, I’m assuming a portion of the people reading this have heard of 32 bit floating but are still unsure about exactly what it is. What are the advantages? What are the disadvantages? When I asked a friend of mine, who is also an experienced engineer about 32 bit floating point, he told me he didn’t know and had never used it.
After having this discussion and then immediately seeing this tweet from accomplished Brooklyn based producer Andrew Maury, I knew I had to finally figure out what the hell it was and if I should be using it.
So, what is 32 bit floating?
The Wikipedia article tells us it’s,
A computer number format that occupies 4 bytes (32 bits) in computer memory and represents a wide dynamic range of values by using a floating point. In IEEE 754-2008 the 32-bit base-2 format is officially referred to as binary32. It was called single in IEEE 754-1985. In older computers, different floating-point formats of 4 bytes were used, e.g., GW-BASIC’s single-precision data type was the 32-bit MBF floating-point format.
Alright, well, that about wraps it up… That was almost too easy. Ha. Ha.
Let’s start with the definition of bit depth because I know that one, and it’s not too difficult to understand. Bit depth is what decides the dynamic range of an audio file.
So 32 bit floating means more dynamic range, right? Not exactly.
So is 32 bit floating better? Higher bitrate means it’s better, right? Sort of.
So it turns out the reason no one knows what 32 bit floating is… is because it’s kind of pointless for most engineers even to bother worrying about it.
A video on the Reaper blog is one of the only sources I found that explained 32 bit floating in a practical way. This explanation is easy for a person that doesn’t like spending his time thinking about digital signal processing anymore than he has to.
So… 32 bit floating is a 24 bit recording with 8 extra bits for volume. Basically, if the audio is rendered within the computer, then 32 bit floating gives you more headroom. Within the computer means things like AudioSuite effects in Pro Tools and printing tracks internally. So say you decide to print a compressor, and the output level is peaking badly… If you are using 32 bit floating, you can bring the level down and restore the headroom so the file won’t be distorted. If you were recording to a tape machine, this wouldn’t be impossible. You can’t just record a bass that’s clipping and restore the headroom afterward. The benefit of 32 bit floating is when processing internally, BUT the downside is the files it creates are 50% larger than standard 24 bit audio files.
Most experienced engineers don’t need to worry about headroom as they probably already know how to make sure levels are never clipping when they aren‘t supposed to be. This article from ask.audio says 32 bit floating will also help reduce unnecessary noise introduced by AudioSuite dithering and rounding errors during signal processing in Pro Tools.
Maybe I’ll write an article in the future where I run some tests to see if there is a noticeable difference between AudioSuite effects processed with 24 bit and ones processed with 32 bit floating.
Update: Most DAWs process in 32 bit floating therefore, if you are processing any audio, it is converted to 32 bit to be processed and then converted back to 24 bit. If conditions permit, it is best to work in 32 bit floating all the way through until mastering to avoid any unnecessary conversion artifacts. Once the project is mastered, you can have the mastering engineer convert the final audio file to whatever sample and bitrate you need.