32-bit Float: Is Clipping a Thing of the Past?

Technology advances, new things come out and some of us jump right onto the new hot thing to stay ahead of the curve before we know what it actually does and how it can help us. HD TV? I didn't need glasses after all! 4K cameras? Goodbye decision fatigue! 32-bit audio? Behold, no more touching those greasy faders! You probably have read or heard a little something about 32-bit float audio already, perhaps about not needing to set levels if you're using it, and I want to ensure that you know what it truly can and cannot do for you. Let me first explain what bits are and the science behind 32-bit float recording, to help you understand what I'll say later in this article.

What is a bit?

A bit is a measure of digital resolution. "Resolution is the number of bits being used to represent the amplitude dimension, the number of steps in the stairs." (Holman, 2012, p. 64) Sixteen-bit provides a dynamic range of nearly 96 deciBels, as each bit gives about 6 dB of dynamic range (16 x 6 = 96 dB). Therefore, industry standard 24-bit provides about 144-144.5 dB of dynamic range, 32-bit fixed integer provides about 192 dB of dynamic range, however, 32-bit float can represent about 1,528 dB of dynamic range! Wow, how many tootsie rolls tall is that?

What is dynamic range?

"The dynamic range of [an audio source such as a microphone] is the "distance" in deciBels from the equivalent noise level to the maximum undistorted [Sound Pressure Level] (SPL). In high quality microphones this amount may range to over 130 dB" (Holman, 2012, p. 82) The challenge is, there are no microphone preamps (found in wireless transmitters, receivers, mixers and recorders) that can handle this much dynamic range. Sound Devices Scorpio, with 32-bit float recording, only has 120 dB dynamic range on the preamps and Lectrosonics' DBSM transmitters only have 112 dB dynamic range. As Holman explains, a responsibility of the sound recordist is to fit this "gallon" of audio output, first into a smaller "quart" jar (preamp such as wireless system, or a front end preamp on a mixer), then into an even smaller "pint" jar, being the recording medium. Increasing the recording bits, increases that pint sized jar, possibly up to the size of the quart jar of the preamp.

Some microphones have a -10 dB pad built in, which gives the sound engineer some control over where the audio is they want to capture within that dynamic range. We must determine how loud is the quietest sound we want captured (background sound) and how loud is the loudest sound we want captured (yelling) and ensure we fit all of that completely within all stages of audio processing. We set the gain on the microphone either with a switch on the mic or, most often, with microphone placement. A skilled boom operator will often "give it air" when talent yells; they'll be as close as possible for whispering, then back off just before a scream, then right back in for whispering. Or for lav mics, we can place the mic a few inches farther from the mouth if we feel we'll need that. Then setting the gain properly on the transmitter and receiver is another level of control to select which small portion of audio we want captured. Next, choosing the preamp gain or trim is yet another level of control. Fourth, adjusting the fader on the mixer or recorder is yet even another level of control. More control helps the sound engineer to record everything we need but if any stage in this process is not properly set, there will either be hissing, due to low noise floor, or distortion due to clipping.

So, increasing bit size from 24-bit (industry standard) to 32-bit, for example, would increase the size of that pint jar to a quart jar, that of the preamp stage before it,

So despite 32-bit float being capable of having 1,532 tootsie rolls, sorry, dB of dynamic range, in this case, it may never actually have more than about 112-120 dB of audio information. Sorry kids, put that Halloween candy back. That's one reason why we need to ask, do we need 32-bit float? We'll continue debating this, because, at this point, 32-bit float audio can be helpful, up to a certain point. 32-bit float audio theoretically, can be helpful to give more headroom to prevent clipping and reduce hiss from the noise-floor.

Do we need 32-bit float recording?

So, with that science and logic, I bring you to my professional opinion on 32-bit float audio and if we need it. If you have a professional audio team, 32-bit float audio has been considered not particularly useful for professionals. 32-bit float audio can help prevent distortion due to clipping in rare cases, but we are trained to not allow this to happen. I've been mixing location sound for film and TV for over 15 years and if my memory serves me well, I haven't had distortion on my recording end since I was in my first year or two of mixing. Like I said,

Sometimes a monsoon comes and all we can do is try to move that mic as far away as quickly as possible. Timing. Let's debate this more. For the rare instance, that a non-audio professional is using a transmitter that is recording (illegal in North America without Zaxcom), eliminating many of the variables in the signal chain discussed earlier, and this transmitter is recording at 32-bit float audio, the idea is that the quart-sized jar of this preamp won't distort now with more dynamic range. However, when there's distortion or clipping, in my experience, it's almost always due to too high of an input going into the microphone.

That is why 32-bit float audio is not necessary. 32-bit float does not fix anything; it's at the end of a signal chain and all it does is help ensure that its own final stage is not a problem, and even that's not a guarantee. Since 24-bit is industry standard and is expected to remain so, the 32-bit audio will then have to be converted down to 24-bit, requiring proper mixing in post anyways, plus 33% larger file sizes.

When to use 32-bit float

I recommend not changing how you already do things, but for some, adding 32-bit float recording can help. If you're already a journalist quickly recording your own news videos and audio yourself, add 32-bit float to your workflow and that can help reduce the amount of distortion you may already be getting or reduce some hiss caused by extremely low recording levels.

Yes, it can undistort that distorted audio caused by poorly set recording levels. But it cannot undistort distorted audio caused by a mic preamp, wireless receiver, transmitter or microphone. If you're a production company with a crew including a camera team, an audio person, etcetera, 32-bit float will be almost completely useless. So, if you think 32-bit float will eliminate the need of an audio professional, you're in for a rude awakening in post-production. You're opening the floodgates for a whole list of issues that 32-bit float or AI cannot fix. Maybe they'll help on 2/10 issues, but an audio professional will ensure you're good on 9 or 10/10 of those possible issues, even without 32-bit float or AI. Let's get it right the first time.

References: Holman, T. (2012). Sound for Film and Television (2nd Ed.) (pp. 64-82). Focal Press.

Brett Ainslie is a NYC based freelance non-union Production Sound Mixer owner/operator.

He has been mixing sound on location for Film & TV since 2010 for narrative feature films, TV commercials, corporate videos, musical and corporate event live streams and broadcasts, digital content, documentaries, sports and network reality shows. Brett has mixed sound for TBS, HBO, Showtime, Bravo, Disney ABC, Discovery, Food Network, Fox, VH1, A&E, ESPN, MTV, National Geographic, Bloomberg, Vice and more.