Recently, I wrote an article on why 32-bit recording is ultimately superfluous for a location sound mixer. Since publishing it, a few things have happened in the audio world.
Rather than edit my original piece, I want to add a few new thoughts on 32-bit recording. If you haven’t read my previous article, you can do so at the link below.
I wrote the last article because I figured people would see the numbers and assume that 32-bit is higher quality than 24-bit since it’s a bigger number. It isn’t. As I mentioned previously, a higher bit rate after a point doesn’t actually sound any better, it just gives you more resolution when you’re applying effects in post. Beyond 24-bit there is no improvement in that regard, though if we’re being totally honest, 20-bit would be more than sufficient, but computers like multiples of 8.
More important to quality is all the stuff in between what you’re recording and the process of writing the files. The quality of the microphone used, which is often measured in maximum sound pressure level (SPL), frequency response, pick up pattern, and qualities that you can’t really assign numbers to, such as warmth. How well does it reject off-axis noise? (Okay, they do measure that in decibels and angles, but your ears are a better judge of that than the spec sheet.)
Then there’s the quality of the preamps, where you can measure things like the noise floor and dynamic range, but otherwise unquantifiable qualities like whether or not they sound natural or warm. How good is the analog to digital converter (ADC)? Then there’s the trim and fader settings before the file is even written. Trying to explain all those numbers would be a mouthful and would ultimately tell you nothing.
In that sense, 32-bit floating point recording looks really good on paper.
The other use is in special effects recording. Say you want to record something really loud, like a lightning strike. Up close, lightning has a sound pressure level of about 120 decibels, whereas rainfall is typically about 50 decibels. So your recording would need a dynamic range in excess of 70 decibels, which is hard to gain stage for. For comparison, the dynamic range of a quiet library (40 decibels) and a person speaking at a normal volume approximately one foot (30 cm) from a microphone (70 decibels) is about 30 decibels. So assuming your preamps and microphones are really good, 32-bit float would help an effects recordist capture the full dynamic range of a lightning storm, assuming your equipment doesn’t act as a lightning rod. Maybe record that lightning storm from a safe distance, a distance where there’s only a 40 decibel dynamic range. Same applies for any sound effect with a high and unpredictable dynamic range. Gunshots. Building demolition. Things of that nature.
It is interesting that it’s becoming the norm on sub-$1,000 prosumer recorders with the Sound Devices Mix-Pre 10 II being the notable exception. Meanwhile, the brand new 833 and Nova ($4,000 and $5,000 respectively) top out at 24-bit. Part of it is professional workflows have been 24-bit for years and there are a lot of older recorders that still sound amazing and are in the field today.
Also, different markets. People like me who record dialogue for a living are more concerned with things like the track count, inputs, power consumption, and ease of gain staging among other things. It’s easy to design a four thousand dollar box that does all that, a lot more, and has really quiet preamps that make 32-bit float completely superfluous. Not so easy when designing a sub thousand dollar recorder for prosumers who don’t do sound for a living.
So in that application, 32-bit does provide a little extra safety at the cost of a little more CPU time and file size, but it won’t sound any better than the more expensive professional offerings. And if you don’t follow this blog, it does sound a lot better on paper.
32-bit recording is a useful tool, but one that only has any kind of advantage in very specific situations. But it’s just that, a tool, and not all tools are necessary for all jobs. I think that’s the best way to think about it for now. Maybe things will change in five or ten years, but for now 24-bit will remain the standard.