[Note: As of this writing, there are two portable location recorders now made with 32-bit float recording capability: The Zoom F6 and the Sound Devices MixPre II series (3, 6, and 10). All of these products are currently available and began delivery in September of this year.]
From the beginning of location audio recording until September 2019, the primary technical challenges to overcome were signal-to-noise ratio and distortion. “Signal” being the original sound entering the microphone, mixer, and recorder. “Noise” being the unwanted additional sound created by the process. “Distortion” being any alteration of the original signal, whether it be additional noise, harmonics, transistor clipping, etc. The method for optimizing signal-to-noise and reducing distortion has always been user application of proper gain structure, meaning that the various links in the audio chain were set to optimum levels — not too high, not too low — and the use of compression/limiting — not too much, not too little.
For production mixers who had to live with a degree of noise and distortion (all of us), the best balance and best way to achieve it was a matter of opinion, making the craft more art than science. But now, at least when it comes to recording original tracks, 32-bit float recording technology makes “proper gain structure” in the recorder an antiquated concept. Now it can be as simple as: 1) Plug in the mic or other source into the recorder; 2) press the red button; 3) repeat.
OK… you don’t believe it, but you will.
A particular problem for recording original tracks, such as dialog for film and video production, has always been that, even with plenty of rehearsal, it is never fully known how loud or soft an actor is going to act. To accommodate this unknown, several techniques are used, each with its own inadequacies. The first of these is input gain, typically set so that the loudest expected sound is just below maximum, while also reducing a few dB more for actors who get excited when “action” is called. Then the fader — if things get too loud, the fader is pulled down to reduce the risk of overload distortion, though this creates its own type of distortion by altering the original dynamics of the actor’s voice. Electronic limiters and compressors are often used to help keep the signal high while also preventing overload, but the very act of limiting and compressing is still a distortion of the original dynamics.
To further increase our odds of delivering the best tracks, “iso” (isolation) tracks are used to record each source on its own track before encountering a fader (“prefader iso”). However, since part of the reason for having prefader iso “safety” tracks is to reduce the chances of overloading surprise peaks, these tracks must be recorded, on average, about 12 dB lower than ideal. Adding to this problem, since these tracks are largely unmonitored, they are often recorded even lower (better too low than risk clipping). With 24-bit fixed files, the necessary low level of isolation tracks means unused headroom, which can result in a noticeably degraded signal when the levels are brought up 15 dB or more (for example) in post-production. All of these challenges and quasi-solutions are handled with a mixture of technical prowess, artistic aptitude, a sense for the tasteful, and a lot of luck, all stirred with the hand of experience. It is an art.
The path to perfection
Analog tape was quite good, but hiss was a real problem. Tape hiss was replaced with 16-bit DAT recorders, which were replaced by 24-bit nonlinear recorders considered to be the ultimate, and still dominate today. However, now that these 24-bit portable recorders allow higher and higher track counts to be used on location, noise and artifacts have, more and more, become a concern again. Also, it is easy and even common to accidentally record 24-bit files too high, creating nasty distortion. Now it appears that 32-bit float recording addresses these concerns once and for all: The only audible limitations of a 32-bit float recording are with the microphone itself or other source going into the recorder. With 32-bit float, regardless of how low the signal is being recorded, no noise or distortion will be added when brought up to normal mixable levels. It also doesn’t matter how high a signal is being recorded,* because a 32-bit float file has headroom that is 770 dB more than a 24-bit fixed file. That’s right: While a 24-bit fixed file has a 0 dBfs limit, a 32-bit float file can be recorded up to +770 dBfs. Not a misprint. That’s 770 dB above 0 dBfs.
32-bit float recording technology allows what seems to be impossible: Recording at a level so low that it is inaudible when monitoring at normal levels and not even visible on the meters, and also allows recording at levels so high they are impossible to achieve. This inaudible, invisible, low-level recording or incredibly high-level recording can then be normalized to full reasonable, workable level (<0 dBfs) with no additional amplifier noise, distortion, or artifacts. This goes against everything preached about the need to optimize gain structure at the recorder, and even seems like the ultimate April Fools joke by an audio manufacturer. But it’s no joke. Information in the links at the end of this article will explain and demonstrate 32-bit float recording in detail.
Utilizing 32-bit float files requires the capability in a playback device or DAW workstation, but it’s already more common than you might expect, and will certainly be everywhere, soon. The 32-bit float recorders by Sound Devices and Zoom recorders can play back and remix their 32-bit float files. Many post-production systems are already 32-bit float capable, including:
Adobe Audition 2015.2.1
Adobe Premiere Pro 2015.4*
Apple FCPX 10.4.6*
Audacity 2.02
iZotope RX7, Pro Tools 12*
Reaper 64 5.979
Steinberg Cuba LE 9.5*
*Waveform displays on these systems can still show as clipped when reduced below 0 dBFS.
**Output on this system can still distort if signal is not reduced below 0 dBFS (MAC OS Mojave)
For more on the technology of 32-bit float recording, the Sound Devices company website has several articles and FAQs here.
Sound Devices also has some excellent extreme demonstrations of these scenarios, comparing 24-bit fixed recording to 32-bit float recording:
Recording past 0 dBFS without clipping
Recording inaudibly low without adding noise
Believe it. You’ll eventually be using it.
-S&P-
GLEN TREW is a Production Sound Mixer with 40 years of experience in film and video production. Based in Nashville, Tennessee, he is the president of Trew Audio (a US and Canadian leader in sales, rental, and service of film and video sound equipment), the president of Remote Audio (a manufacturer of specialty audio products for film and video), and a member of the Hollywood IATSE Union Local 695 (active, retired).
Images courtesy of Sound Devices