Ah 32-Bit!

The savior, the technology that will prevent you from ever losing a recording ever again. Right?

It’s time to chat about 32-Bit recording. And not the GOOD stuff about it. There is good stuff. 32-bit Analog to Digital converters offer an insane level of headroom. It is literally impossible to clip a 32-Bit converter. Or rather, the sound needed to clip one would have to be around 1,500dB. What this means in practicality is that you never need to worry about clipping your A to D converter. You don’t even really have to worry about setting levels, as 32-bit can also record sound at such low levels and at such resolution that you can simply gain them up a ridiculous amount and still maintain full quality. I’ve recorded sounds that didn’t even show up on the meter and was able to gain it up 70dB or more and it was without issue. Some of the new breed of recorders such as the F3 from Zoom don’t even have record level functionality. The idea, you just set your gain in post.

But. There are some ‘dark’ truths about 32-Bit that are often overlooked, and in this post, I want to have a chat about at least a few of them.

The first is this myth that with a 32-Bit recorder, you can never ruin a take, that you don’t need to be concerned about levels, and that you’ll be able to recover from anything. This is not the case.

Think of the chain of your sound through your recording gear. First there is the microphone, then the preamp, and THEN the A to D converter. That leaves two vulnerable stages before the ADC.

You can still clip the microphone. And you can still very easily clip the Microphone Preamp. In fact, the Zoom F3, a wonderful Swiss-army knife of recorder, has a notoriously LOW preamp headroom. About 10dB lower than the headroom offered on the Sound Devices Mix-Pre II series recorders. There have been many times that I have seen the dreaded ‘Input Level Exceeded’ message flash across my F3. It is a bit shocking to me that Zoom would tout a 32-bit recorder with a preamp headroom low enough that it routinely causes problems, and not even on impressively loud sounds.

And of course, you can just as easily still exceed the volume capability of your microphones. Many microphones can record exceptionally loud sounds, dynamic mics for example can easily record 140dB or more. But many condenser microphones have max SPLs of around 120, which is quite easy to exceed. When you ‘crunch’ a microphone by overloading it, nothing is going to save you.

32-Bits will not save you in either of these situations. Knowing about microphones, understanding the need to ‘pad’ a microphone to avoid overloading the preamp, these things are still necessary skills to have. That is one issue I have with 32-Bits, it seems to want to say to a new recordist, “Don’t worry, I’ve got you, no need to pay attention.”

Another issue that I see (luckily not SO widespread, yet) is the use of 32-Bit files as a ‘release’ format. I see several vendors selling 32-Bit files. This makes literally zero sense. Here’s why. When digital audio came of age, everything operated in 16-Bits. This is the resolution of a compact disc. 16-Bit technology was (and is) still amazing. It offers up a dynamic range of 96dB, quite a lot in fact, and is capable of very high-quality audio. But it did have some limitations, particularly its ability to record low level signals with decent fidelity. So along came 24-Bit, which has sort of become the ‘new’ 16-Bit. Most feature film work is done at 24-Bit, and a lot of music is now released at this bit depth. 24-Bit has a noticeable improvement over 16-Bit, especially for its ability to record low level signals with sufficient resolution. It also increases the headroom to a very usable and impressive 144dB. You will hard-pressed to find many sources in nature that exceed 144dB. For example, that’s louder than a jumbo jet flying 100 feet over your head, louder than the loudest rock concerts. But 24-Bit when recording, paired with around 60dB of gain that many recorders have, meant it was easy to clip the ADC if you had your level set high, and a loud event occurred. Once clipped, your only hope was for something like Izotope RX to try and repair the damage.

Along comes 32-Bit to the rescue. We should be clear too, that when we say 32-Bit, we actually mean 32-Bit float. I won’t try and explain the difference as I’m not qualified to, but as I understand it, 32-Bit float is basically a moving 24-bit range of precision within a 32-bit space. But anyway, just know we when talk about 32-bit in audio we almost always mean 32-bit float. There is such a thing as 32-bit fixed, and it would have a dynamic range of 192dB. It’s the ‘floating’ part of 32-bit float that gets you the insane 1,500dB dynamic range we are talking about here. But it also means that you are basically still recording with 24-bit ‘resolution’. A 32-bit float and a 24-bit recording at the same gain should sound the same.

Now 32-Bit Floating math has been used in technology for quite a while. Your internal mix engine inside your DAW most likely uses all 32-Bit (or higher) math for summing and processing of audio signals. It does this to never risk clipping the sounds as it does all its magic summing math. And when 32-Bits was released as a recording technology, the selling point was that you’d never have to risk clipping your recording either. And again, aside from the microphone and preamp pitfalls, this is true. With 1,5000dB of total dynamic range, a 32-Bit ADC converter is un-clippable.

But what happens next. You have all these 32-Bit files, what to do with them now?

Well, what you SHOULD do, is to master them into 24-Bit files. Why? Because 24-Bit still has so much dynamic range and headroom that it can hold anything you need it to. You should load your 32-Bit files INTO a 24-Bit session. ProTools, and likely all other DAWS, will not need to convert a 32-bit file to load into a 24 bit session. What this still allows you to do, is to take that 32-bit file, and normalize, or gain them, ‘into’ a 24-Bit range. This should be done as part of the mastering process (and you should of course master every file you record!).

What some people are doing is to master in 32-bit and release libraries in 32-bit. Why? I suppose they presume either that it’s ‘better’ sound quality, or they recognize people’s ability to be easily duped by ‘higher numbers’. Remember the megapixel wars in both TV and Cameras. But it’s silly for two reasons. First is, it’s just wasted space. A 32-bit file takes up 33% more room than a 24-bit file, and since if the files were mastered properly, you get nothing for that wasted space. Second, some programs may not like 32-bit files. And if you say take a 32-bit file that DOES have dynamic content over the 24-bit threshold of 144dB, converting that file into 24-bit without adjusting may get you a clipped 24-bit file.

And many designers choose to maintain a library of sounds in FLAC (Free-Lossless-Audio-Codec). Sadly, for now FLAC does not support 32-bit files.

There is one last issue with 32-bit that can be a bit nasty and is being talked about more and more. This has nothing to do with 32-bit per se, but in the current IMPLEMENTATION of it. Because quite a few manufacturers (including Zoom and Sound Devices) moved quickly into 32-bit recorders, and I suppose because there was no cost effective true 32-bit A-D converters available, a new Dual A-D setup was put in place. There may be triple-ADC setups out there I’m unaware of). While this gets technical and honestly over my head, basically the idea is this. Instead of a single A-D converter with a very wide dynamic range, you use two (or more), each tuned to a different gain range. You have a low-gain and a high-gain converter, and the recorder will automatically pick one (and even switches between them while recording) as it needs. This combination of converters will provide the higher dynamic range.

The issue is this. The high-gain A-D has significantly higher self-noise than the low-gain A-D, usually in the ultrasonic range. And what this can mean is that if the A-Ds switch in the middle of a take (which they can easily do) you can be left with an ultra-sonic burst of noise, as the higher noise floor A-D switches on. Now that noise is likely well above the level of human hearing, and you shouldn’t notice it until and unless you decide to pitch your sound down. At this point, especially with extreme pitching, that noise burst drops into the audible pitch range.

This sound has been referred to as ‘tearing’ and quite a few people have noticed it, I personally have not experienced this in a way that has been detrimental, but many others have. This issue for sure exists on Zoom 32-bit recorders like the F3 and F6 (and I assume on the new H1E, H4E and H6E as well). I’m told by someone in the know that Sound Devices claim to have somehow fixed this issue via firmware to eliminate this issue on that their second generation MixPre series. But they will not reveal ‘what’ they are doing to surmount this. Since I mostly use my high-sampling rate microphones like my Sunken CO-100ks with my MixPres, perhaps that is why I haven’t encountered this before. But on Zoom, it is a real issue that may affect some.

So where does that leave us. 32-Bit is still great. It can provide a lifeline in certain situations. I think it’s a very useful technology for say a production sound recordist, who can’t risk losing a take when an actor suddenly decides to shout out a line of dialog. Or in situations of say recording a thunderstorm. But people should understand the risks.

So, in conclusion and quick summary:

1. 32-Bit will not save you from clipping your microphone or your microphone preamps. You need to understand this, and you need to understand mic selection, padding, preamps etc.
2. 32-Bit is a great acquisition format and a poor ‘release’ format (you are simply wasting hard drive space with no other gains)
3. Be aware of the implementation of 32-bit in Zoom recorders, especially if you plan on pitch-shifting down the sounds you will be recording, as the ultrasonic tearing issue is real and could affect you.

Happy recording everyone!