The new Zoom F6 MultiTrack Recorder was the first audio recorder to hit the market that supports 32-bit Floating Point audio recording.
For this Audio Assist column, we’ll explore a relatively new technology in location-sound recording called 32-bit float point audio, which is a new way of recording sound and processing it in post.
Characterizing Common Recording Problems
Before delving into this new technology, I’d like to discuss some common problems you might encounter when recording audio. (That’s because one of the main benefits of 32-bit float point audio is to overcome certain audio recording problems.)
Up until now, when it came to recording audio, there’s an axiom often used by many who work in production: If you capture bad quality audio, generally you’re out of luck. It’s a statement that seems odd in an era when there are so many great postproduction sound tools, plug-ins and technologies.
However, what exactly does this axiom mean? What is actually meant by “poorly recorded location sound?” To me, it can mean many things. Just consider the following list I’ve compiled of recording problems due to poorly recorded location sound and their effects on the audio file to see what I mean:
- Incorrect Microphone Placement: Low signal-to-noise ratio (S/N), echo, reverb, too much ambient noise, clothing/facial hair/jewelry rubbing on lavalier, etc.
- Ambient Noise Issues: Sound recorded in any environment with your dialog that is not the desired sound—wind, AC hum, air conditioning/heating noise, elevators, air or street traffic, lawn mower, leaf blower, dogs barking, birds chirping, construction noise, trash trucks, insects buzzing, etc.
- Excess Reverb/Echo: The result of recording talent in a reverberant room or environment.
- Bad Microphone Choice: Each type, model and brand of microphone reproduces a voice or sound in its own unique way.
- Defective Equipment Noise or Bad-Quality Recording: Such unwanted audio can be due to a variety of causes, including cable noise, intermittent or shorting connection points with the gear and or cables, RF noise issues with wireless microphone systems, moisture/humidity levels with certain microphones, dirty pots on recorders or mixers, etc.
However, the ones I just listed aren’t really the most common problems. In my experience, the two most common are recording levels that are either too low or too high.
Recording Levels That Are Too Low
If you record your audio signal at too low of a level, here’s what will happen:
You—or the video editor or sound mixer (whoever is responsible for the soundtrack of your project)—will have to raise the audio levels to a standardized level in post to distribute the edited project. But when you raise those low audio levels from a very minimal nominal level to a much higher, corrected level of, say, -12dB or -6dB peak, you also increase all of the corresponding noise in the recording.
The end result is that the increased noise will often mask the dialog that you need to hear. So, the only choice is to raise the nominal audio levels and apply judicious dialog editing and noise reduction.
Modern noise reduction tools are amazing and have improved tremendously over the past decade, but even the best of them tend to noticeably alter the tone and quality of the remaining audio that you want. I use an audio repair toolkit called iZotope RX 7. It’s a good tool. But it can’t perform miracles. When you apply software like iZotope to a low-level recording, the end result is louder and somewhat cleaner, but the voices tend to sound metallic, with compressed dynamic range and often a honky-mid-range enhanced sound that isn’t natural.
Recording Levels That Are Too High
Conversely, if you record your audio levels too high, you’ll over-modulate—or “distort” or “clip”—the audio waveform. In my experience, this can happen even with the most competent of location sound recorders. If you set your recording levels on your talent and they are changing the level of their voice in the performance, generally the sound mixer will attempt to ride the gain levels on the recording device, whether the sound recorder is a camera, sound recorder or portable recorder placed on the talent.
But if the talent laughs, screams, yells or makes any kind of loud sound, the resulting recording will often be distorted.
A commonly used strategy to mitigate this problem is to use a duplicate input second channel and set the recording levels on the second channel a pre-determined amount lower than the main recording channel. This is often referred to as a “safety” channel and can help to reduce the over-modulation or distortion. But then you have to dedicate two channels/tracks on your recorder to each microphone.
Often in scenes with more than one or two talent, you run out of channels to record each talent with a safety channel.
The Benefits Of 32-Bit Floating Point Audio
However, when it comes to the two most common recording problems—recording levels that are either too low or too high—there is a powerful solution: 32-bit floating point audio. And here’s why—it has greater dynamic range than either 24- or 16-bit audio files.
Now, it’s important to point out that at this moment, in 2020, the most widely-used file format in the pro audio world is a 24-bit, 48 kHz WAV file, which has a dynamic range of 144.5 dB. (By comparison, a 16-bit file has a dynamic range of 96.3 dB.)
But the dynamic range that can be represented by a 32-bit (floating point) file is 1,528 dB, which seems unreal when you consider that the greatest difference in sound pressure on earth can theoretically be only about 210 dB—everything from the extreme quiet of an anechoic chamber to a massive shockwave. This means that recording even gunshots, explosions, the loudest screams to the faintest whisper can all be recorded at the “correct” audio level, regardless of where you have set your gain levels on your recorder!
It almost defies logic for how most of us were taught to record audio levels correctly.
Here’s how it generally works: When a DAW or audio editor first reads a 32-bit file, signals that are greater than 0 dBFS may first appear clipped since, by default, files are read in with 0 dB of gain applied. But that’s just how it looks! When you apply attenuation to the file in the DAW or editor, signals above 0 dBFS can be brought below 0 dBFS, undistorted, and used just like any 24- or 16-bit file.
In short, for 32-bit float recording, the exact setting of the trim and fader gain while recording is no longer a worry, from a fidelity standpoint. The recorded levels may appear to be either very low or very high while recording, but they can easily be scaled after recording by the DAW software with no additional noise or distortion.
Limitations And Drawbacks Of 32-Bit Floating Point Audio
While 32-bit floating point audio is very impressive in terms of fixing audio with levels too high or too low, it can’t fix all problems or issues. Here are some instances it won’t repair:
Microphone Distortion: Even though audio recorders that utilize this technology can theoretically record basically any sound without distortion, it’s important to remember that sound enters a recorder through a microphone. And any mic can distort. In fact, it’s still relatively easy to distort the diaphragm of a microphone.
If you are recording exceptionally loud sounds, you still need to choose the right type and specification of the microphone carefully. It’s also important to know where you place the microphone in relation to the sound so that it’s capable of actually capturing the sounds you want to without distorting the microphone’s diaphragm itself.
It Doesn’t Apply To Outputs: One common question regarding this new technology is, “Does 32-bit floating point output from a recorder/mixer mean that the audio signal coming into my camera from the mixer/recorder is also 32-bit floating point?”
In a nutshell, no. If you utilize a 32-bit floating point-capable recorder/mixer and run the output to any camera, the audio that the camera records will be affected by your gain choices, and you will record your audio either too low or too high. The only way to solve this problem is to take the 32-bit floating point files from the audio recorder and set their levels in a DAW or editing app that supports the same 32-bit floating point audio and import them into the video editor or use the video editor itself if it supports 32-bit floating point files natively.
Everything Else: Using 32-bit floating point audio won’t prevent any of the audio problems we listed at the beginning of the article—for instance, if a lavalier is placed on talent and rubs on skin or material or if the lavalier has cable noise, RF noise or shorts. Or, say there are ambient sound issues or you made a poor microphone choice or didn’t place it correctly—32-bit floating point does nothing to save, fix or repair these and many other issues. So, you still need to apply skill and experience to every other aspect of the location sound challenge.
Increased File Size: One downside of 32-bit floating point audio is that the file sizes are 100% larger than 16-bit files and 50% larger than 24-bit files. So, be sure to note any long takes for your project that you plan to record using 32-bit floating point technology.
Recorders, DAWs & Video Editors Supporting 32-Bit Floating Audio
Although 32-bit floating audio is still a new technology, there are already some products on the market that support this new specification. At press time, I found the following audio mixer/recorders available, although most professional audio recorders that are designed from here on out will most likely support 32-bit floating point audio.
Additionally, most modern Digital Audio Workstations (DAWs) and some video-editing software can read 32-bit float files. Here’s what’s currently available:
Audio Mixers & Recorders
- Zoom F6 Six Input, 14-Track MultiTrack Field Recorder
- Sound Devices MixPre-3 II 3-preamp 5-track recorder
- Sound Devices MixPre-6 II 4-preamp 8-track recorder
- Sound Devices Mix-Pre-10 II 8-preamp 12-track recorder
- Tentacle Sync Track E Recorder (announced shipping as of NAB 2020)
Digital Audio Tools
- Adobe Audition 2015.2.1
- Audacity 2.02
- Boom Recorder 8.7.3
- iZotope RX 7
- Pro Tools HD 10 (Pro Tools 2019.12)
- Reaper 64 v6.01
- Steinberg Cubase LE 9.5
- Steinberg Nuendo 10.2
Video Editing Software
- Apple FCPX 10.4.6 – Yes
- Adobe Premiere Pro 2020 (partial support: waveform displays and outputs appear to be 24-bit, outputs will clip if signal is > 0dBFS)
At press time, AVID Media Composer 2020 and DaVinci Resolve 16 didn’t yet support 32-bit float files, but keep in mind that 32-bit floating point recorders only just hit the market last year. That’s why more digital audio tools (like DAWs) support it than video editing tools, as of today. Also keep in mind that you can record 32-bit floating point audio files and just use an audio tool to export “rescued” audio files when and where needed to import into your video editing software.
Lastly, it’s very likely that upcoming updates from all of the popular video-editing applications will soon integrate support for 32-bit floating point audio.
Final Thoughts On 32-Bit Floating Point Audio
We should be excited by 32-bit floating point audio recording because it represents a totally new way to think about audio levels during recording. To be practical, though, as of today, there are only four professional audio recorder/mixers that implement this technology. If you need a new audio recorder/mixer, I highly advise at least trying it out before you buy to see if this is something that is important to your workflow.
Then, ask yourself: Do you shoot, then deliver footage and sound files at the end of a shoot to a client, agency or editor? Are they utilizing a workflow and software that support 32-bit floating point audio? If so, will they be expecting 32-bit floating point audio files?
If you are your own editor or sound mixer, are you using software that supports it? Keep in mind that all of the recorders currently available also can be used for regular 24-bit fixed point recording, so it may be that for your normal, day-to-day workflow, utilizing 24-bit is better. Reserving the 32-bit floating point audio for the times when you may need its capabilities is more practical, at this point.
If you are a solo video shooter, a sound mixer who sometimes has to do bag drops where you leave your sound bag unattended to record, for instance, for car scenes where you cannot be in the back seat to monitor, this technology is great and will help you do a better job.
If you’re a sound designer or someone who records sound design elements, 32-bit floating point audio will change how you record your work. Recording extremely loud or dynamic sounds will become easier, although microphone selection and gain staging in certain scenarios will still be important.
It seems that with the advent of this technology, there is a good chance that future recorders and even camera audio may eventually implement 32-bit floating point audio.
But for now, you’ll have to determine if this is technology that you want to buy and begin using today, or if it’s an exciting preview of what may become commonplace soon. Either way, it’s an amazing tool that holds great promise to increase the quality of audio in many situations.
But remember that this technology doesn’t in any way negate how important it is to hire a professional sound mixer whenever you have the budget to do so as recording good-quality sound is reliant on so much more than just setting correct recording levels. In my view, a sound pro is worth his or her weight in gold. But for those situations where it’s not possible, 32-bit floating point audio could be a key technology for you to investigate.