Float 32 editing

Float 32 allows about 1,500db of DR...but Resolve will only allow 30. Can we please please please unlock the slider limitations? Float 32 is not new so it's a head scratcher that Resolve still doesn't know how to use it.

There’s another thread discussing it.
Hopefully they’ll add it.

I'm confused.

Color is already processed in 32 bit flat.

And we can import and edit 32 bit float audio.

What does this request apply to?

What does this request apply to?

The ability to manage volume adjustments that far exceed the current 30dB limitations set based on the expectance that audio is stored in integer and levels are already 'normalized' to the usable range.

The ability to manage volume adjustments that far exceed the current 30dB limitations set based on the expectance that audio is stored in integer and levels are already 'normalized' to the usable range.

Don't forget that cranking up a signal past 120dB starts hitting the threshold of pain...



You can already crank up signals by 30dB right now in Resolve, and I'd suggest if you need to turn a track up more than 30dB, something is very wrong in your signal chain. Even if you did crank it up 60 or 70 or even 80dB, don't forget that the microphone self-noise and the preamp noise also gets turned up, so the end result is you get a noisy signal. This is the reason why sound mixers on TV and film sets are employed.

Don't forget that cranking up a signal past 120dB starts hitting the threshold of pain...

I'm perplexed as to why you are still talking in context of thinking that managing levels past 30dB is not necessary. You seem to still don't understand the concept and practical application of 32bit float recording. I won't even try to explain since you've been told so many times already in the bigger thread that was going on about this topic.

Don't forget that cranking up a signal past 120dB starts hitting the threshold of pain...

dB in a digital editor is not the same thing as dBSPL. In a digital system the upper limit is 0dBFS because going over this will clip your signal, but that has nothing to do with dBSPL. dBSPL is used for the sound produced by your speakers/headphones. Also, how much you can boost before clipping digitally depends on the level of the original signal. If you have a -121dBFS signal, you can crank it up by 120dB and still be under the 0dBFS limit without clipping.

Even if you did crank it up 60 or 70 or even 80dB, don't forget that the microphone self-noise and the preamp noise also gets turned up, so the end result is you get a noisy signal.

If I turn a clip down by 80dB and render out a wav file in 32bit float, I could sent that 32bit file to you and you could boost it by 80dB without introducing any more noise than the original I have here.

Don't forget that cranking up a signal past 120dB starts hitting the threshold of pain...

dB in a digital editor is not the same thing as dBSPL.

I know, but my point is the same. Get it right on set and none of this is necessary. All you're doing when you record actors at low levels on set is you delay those decisions until post, and dump it on the head of the poor editor... which I think is wrong. To me, you need usable sound at the right level, even in dailies. it's not that hard to do. (I would compare it to people who shoot and don't bother adjusting exposure correctly: there are limits in terms of how far things can be cranked up when they start out way too low.)

For more on the question of recording with 32-bit audio bit-depth, I refer you to this recent discussion from more than a dozen Hollywood sound mixers:

https://jwsoundgroup.net/index.php?/top ... ent-419927

I had a conversation with an Emmy-award winning re-recording mixer just last weekend and asked him the 32-bit question, and he said the exact same thing I did. He also volunteered: "I'm baffled as to why people in film school aren't being taught that 24-bit audio already has 144dB of dynamic range, and it's important to capture as loud a signal as possible without distortion -- at least at -10dBfs. But recording at (say) -70 or -80, which would require turning up the signal 50dBfs or 60dBfs, is completely unreasonable."

I brought up the mic noise and preamp noise issues, and he pointed out that the problem is exacerbated when you have (say) 5 or 6 or 7 or more simultaneous microphones in the same scene. And also agreed that nobody in town is mixing in 32 bits... all the stages (and delivery mechanisms) are set up for 24 bits. This mixer (who has far more experience than I do in the sound business) said he was perplexed and disappointed by the lack of sound knowledge by filmmakers, particularly when all this information is readily available if you just look for it. One Google search will give you the information on how to set mic levels and what gain staging is. Both are crucial concepts in sound.

Don't forget that cranking up a signal past 120dB starts hitting the threshold of pain...

dB in a digital editor is not the same thing as dBSPL.

I know, but my point is the same.

Ok, but your point about cranking up sound in a DAW will reach the threshold of pain was incorrect. Also, your point about cranking up the volume by 80dB will result in a noisy signal was incorrect as well. I gave an example to show this, which was turning down a signal by 80dB, rendering it out as a 32bit float and then turning it back up by 80dB again. This will not introduce more noise, so boosting by 80dB does not guarantee high noise.

I brought up the mic noise and preamp noise issues, and he pointed out that the problem is exacerbated when you have (say) 5 or 6 or 7 or more simultaneous microphones in the same scene.

You seem to be assuming that 32bit recorders use typical preamp and converter designs, which they don't. I imagine the exact design varies from product to product, but from what I've seen they use some combination of multiple preamps and converters then switch between them depending on the incoming signal level. Then these signals are recombined back together and stored in a single 32bit float file. They are designed to capture a wide dynamic range without introducing noise.

In fact, while looking up a video to post for you, I saw that the Zoom F6 32bit recorder doesn't even allow you to adjust the preamp gain. It is locked in at whatever level they set it to. That should tell you how much of a non-issue preamp noise is for 32bit recorders.

Here is a good audio demonstration (starts at 3:00) so you can hear it for yourself. They compare 24bit and 32bit recordings by going from a barely audible whisper to screaming and clipping the input without changing the recording level. Then in post they re-balanced it. The 24bit recording sounds terrible in both cases, but the 32bit recovered both the whisper and screaming parts just fine with no noise or distortion.

I had a conversation with an Emmy-award winning re-recording mixer...

And also agreed that nobody in town is mixing in 32 bits... all the stages (and delivery mechanisms) are set up for 24 bits. This mixer (who has far more experience than I do in the sound business) said he was perplexed and disappointed by the lack of sound knowledge by filmmakers,...

Emmy award winning or not it doesn't make any sense. Nobody mixing in 32bits? Pretty much all DAWs are 32bit float internally. If not 64.... Only mastered deliverables are down to 24int or 16int. This is no different than mastering/grading film. Or would you like to make the argument that our color tools and CG renders don't need to be 16/32float either because there is already enough dynamic range for the deliverable?

The idea that you just take his/her word for it because you expect they know more than 'filmmakers' rather than reading and actually understanding the topic is weird. On top of that trying to defend against it in the form of being against BMD developing tools to work with such data and telling people "just record properly" is condescending and childish.

Perhaps this attempt can make you understand the principle better. Here's an ARRI sample footage frame in AP0 linear exr. Boost the exposure by 10 stops and tell me if there suddenly is more noise or your eyes start to bleed because of the boost.
https://www.dropbox.com/scl/fi/o5y4zycc ... 16pzr&dl=1

Wasn't there a full discussion on the subject on a previous FR for 32-bit audio?

Edit : viewtopic.php?f=33&t=145749 << this thread is full of information. I don't get why some people don't see how crazy useful it is. I don't work with 32 bit audio myself, but I would like to have the option if I do.

Marc continues to display his lack of knowledge of the reason for 32 bit float audio. Until he understands the technology and purpose I think it is pointless responding other than to ignore his input maybe on other matters too since this demonstrates his lack of up to date technical knowledge. Maybe his friends in the audio industry do not understand the technology they are using either.

He clearly does not understand the issues of recording as a " one man band " in an uncontrolled environment. If he tries this he may see the value in 32 bit float audio and maybe even things like auto focus etc. My F3 has made life a lot less stressful. Plug in mic, set the display to look like a want ( this is important for further editing ) and press record. Done. No levels that are too low and no clipping. Most of my recording are in reasonable range so dynamics in Fairlight are adequate to get the output I want. However I do understand the need for some applications to get more range.

Get it right on set and none of this is necessary.

Come on.

First things first, the folks "on set" are just a small chunk of the user base. But, even if I'm not in the industry, I know there are loads of pros who swear by 32-bit audio, for good reasons. Plus, it virtually costs nothing, considering how cheap disk space and processing power have become.

And time matters,...a lot. Whether you're a pro or just an amateur shooting videos in your garage, time is precious. Let's not forget about health too. Reducing frustration, anger, stress, sleep... it all adds up in a person's life. And all for something that's almost cost-free.

It's like insisting on shooting JPEG instead of RAW (or RAW+JPEG). Having every bit of data ready to be twisted appart for future needs is a no-brainer, whether to fix errors, for convenience, or for creativity purpose. Shooting in RAW doesn't mean you can't just shoot JPEG.

Or think about telling someone wearing a seatbelt while driving: "Why bother with a seatbelt?! Just drive cautiously and keep an eye out for other drivers."

When I was doing a lot more photography, I always hated people arguing against shooting RAW with the same argument : You should get it right in camera - Or : ALWAYS shoot manually, don't even use aperture priority - Or : It's okay to have "grain" (then it's just disgusting digital noise), and if you should expose your scene properly (as if every single real life situation is a perfect situation) - It's the same type of discussion here.

I personally always try to get it right in camera the first time just because I want to see 80-90% of the finish "product" right away.

You can stick to a strict way of doing things, even in any kind of "sandbox."

Reminding the importance of nailing it 'right' (or pretty close) initially is a good thing, don't get me wrong. However, in creative digital space, it's merely a broad, solid principle to keep in mind. And yes, beautiful and/or crazy things done with less assistance will always be more interesting to see (and impressive), but that's another subject.

In other area of life, yeah, most of the time it's imperative to follow rules to get things done properly.

Anyway, you're arguing against points that aren't made. And I think everyone already knows what you're saying.

...tell me if there suddenly is more noise or your eyes start to bleed because of the boost.

What does this request apply to?

The ability to manage volume adjustments that far exceed the current 30dB limitations

Ahhhh.

Already have that request.

viewtopic.php?f=33&t=177369#unread

Reminding the importance of nailing it 'right' (or pretty close) initially is a good thing, don't get me wrong. However, in creative digital space, it's merely a broad, solid principle to keep in mind. And yes, beautiful and/or crazy things done with less assistance will always be more interesting to see (and impressive), but that's another subject.

I dunno. I see mindless arguments like this being made all the time, and it's wearying. These are problems that were solved years ago. You basically have to be smart in production and use good judgement. Coming into a mix with dialogue 30-40dB too low is problematic at best and disastrous at worst.

BTW, my old pal Bob Katz -- one of the best music mastering peope in the world -- pointed out to me today a great page of math detailing what's involved in comparisons between record levels, actual volume, percentage changes, and measurements. It's a sobering document:

http://www.sengpielaudio.com/calculator-FactorRatioLevelDecibel.htm

A lot of this applies more to music than dialogue, but John Purcell's book on dialogue editing (and recording) that I referred to above DOES specifically deal with why it's best to get the microphone levels adjusted correctly in the signal path, which includes the mic, the preamp, the cable, and the input signal going into the recorder. Recording at 32 bits or 48 bit or 64 bits can't salvage bad sound. If you record 100dB too low, cranking it up 100dB will not restore it back to normal, because digital is imperfect: you merely magnify the flaws that are always there.

I'm often saddened by the number of neophytes and students who jump into filmmaking with bad habits and bad assumptions, without any evidence of what they're trying to do. Again: the information is out there on the net and I'm not the first person who's said it. These basic principles go back decades, even to the dawn of digital in the 1980s. But do it your way -- if you're lucky, you'll eventually see (and hear) the folly of what you're trying to do and understand why the math doesn't work. (And similar principles on keeping signals clean also apply to picture and other aspects of post production.)

Coming into a mix with dialogue 30-40dB too low is problematic at best and disastrous at worst.

Did you listen to the example I posted for you before? Can you explain what was problematic about their 32bit audio examples?

As I said before, you are still thinking of 32bit recorders as a single preamp and converter, when that is not how they are designed. They have multiple preamps and converters internally so they can record very low to very high levels automatically without you having to ride the gain. What you have in your head as recording "too low" is in fact well within range for 32bit recorders. The audio example I posted for you before definitively proves this.

You cannot approach digital recording in the same way you would analog recording because the technology is different. Likewise, you cannot approach 32bit recording in the same way you would 24bit recording because the technology is different.

If you record 100dB too low, cranking it up 100dB will not restore it back to normal, because digital is imperfect: you merely magnify the flaws that are always there.

That is also incorrect. You can prove this yourself with a null test:
1) Load any audio file, turn it down by 100dB and then render it out as a 32bit float wav file.
2) Load the 32bit file in on a new track and boost it by 100dB.
3) Invert the polarity on either track and play them both at the same time. If the two tracks are identical then they will perfectly cancel each other out and you will get no reading on the meter.

It should only take you a minute or so to try this test. Let me know your results. I even did it myself and the meter in Reaper showed -INF. That means they are identical.

Did you listen to the example I posted for you before? Can you explain what was problematic about their 32bit audio examples?

3-time Grammy-winning engineer Bob Katz just responded to me on Facebook an hour ago, and minced no words. Bob had this to say about the fallacy of 32-bit recording:

32 bit float direct recording is a SCAM. It doesn’t give you infinite headroom or any better signal to noise ratio. You’re always limited by the headroom of the preamp.

I think it was Zoom that started promoting this “record to float and you won’t ever have to set your gain” bull crap. The recorders with that feature were designed to deal with reporters with no engineering knowledge. So they drop the input gain 10 or 20 dB to reduce the chance of preamp overload under normal conditions. Then they record to float so the reporter doesn’t even have to worry about record level.

The first sacrifice is 20 dB loss of signal to noise ratio to make the system easier to use for the novice. The second sacrifice is the user has no idea what to do with the floating point file

Bottom line as I said: it’s a scam. You don’t get something for nothing.

--------------------------------------------------------------------------------------

Again, these are Bob Katz' words. He's the author of Mastering Audio: The Art and the Science (3 editions), which is basically the last word on the subject, and also several other books, and has been a top mastering guy for at least 50 years that I know of. Bob has extraordinary ears and amazing talent, and is far smarter than almost anybody I know in terms of audio and sound theory. His work is beyond reproach.

You can check out Bob Katz' website here: https://www.digido.com/

Did you listen to the example I posted for you before? Can you explain what was problematic about their 32bit audio examples?

3-time Grammy-winning engineer Bob Katz just responded to me on Facebook an hour ago, and minced no words. Bob had this to say about the fallacy of 32-bit recording:

I take that as a no, you didn't listen to the audio example. Ok, here it is again. Jump to 3:00 for the audio comparison between 24bit and 32bit recordings.

After you listen to it, tell me you don't hear any difference between the 24bit and 32bit recordings.

The recorders with that feature were designed to deal with reporters with no engineering knowledge. So they drop the input gain 10 or 20 dB to reduce the chance of preamp overload under normal conditions.

No, that is literally not how they work. They use MULTIPLE preamps and converters internally in order to record different input signal levels without noise. I have explained this to you twice now, but if you won't listen to me, here is a diagram from Rode:

Rode 32bit Recorder.png (95.42 KiB) Viewed 4778 times

And here is the full video I took that screenshot from:

Are you saying that the companies don't know how they designed their own products and that the specs they have put out are wrong?

If you want to read up on the headroom of 32 bit float, you can find it here.

https://www.sounddevices.com/32-bit-flo ... explained/

Yes, you can clip the microphone signal/receiver if you use a lav BEFORE the recorder. However if the input signal to the 32 bit float recorder is not clipped, you don't need to worry about clipping in the recorder. Obviously try to set your levels properly as its easier to mix on set, but no worries if talent randomly starts yelling or some pyro goes off, just reduce the gain in post and youre all set.

People.... Marc is not prepared to learn anything new or admit he is wrong or doesn't know something. His lack of knowledge of 32 bit float and its intended use likely extends to other topics using new technology too. I will try not to take the bait in future.

Bob has extraordinary ears and amazing talent

He's also working at a much higher level than I am.

For my corporate/event projects, 32 bit Float is akin to shooting RAW video, with results that are perfectly in line with the quality expectations of my clients.

For the doubters, it's true that under normal circumstances, a 32 bit float recording is no better than 24 bit recording, just a lot bigger. But that's not the issue.

In a worst case of 32 bit float recording (levels way too low) the recording is still sufficiently above the noise floor that significantly raising the volume produces little or no additional noise or distortion -- unlike 24 bit recording.

Similarly, no matter how high the levels, you will not clip. This trick is possible in parts thanks to two separate processing chains. Tascam illustrates it clearly:

Capture1.JPG (60.32 KiB) Viewed 4720 times

Capture2.JPG (66.69 KiB) Viewed 4720 times

This may be of no use in well-controlled studio conditions, but that's not for whom this technology was designed.

This is in some manner similar to the shooting raw or log or ProRes discussions.

And it boils down to time, really. Like many others, on personal things or projects or testing, where I want or need ultimate results, I shoot braw.

But for many things, it is simply faster on the back end to shoot tightly in camera to ProRes and get the job out the door.

On pro stills, I shot raw nearly all the time for studio individual clients. For "bulk" jobs, schools churches and such, tightly controlled jpeg.

Because ... time.

Like Marc, I've seen discussions by high-end audio pros who recognize that 32-bit float audio can be "safer" at times, by a small amount.

But they also note that it costs their clients more time in audio post. Again ... time.

My H6 can record 2 files at different gain, one set for expected and tested levels, the other lower for unexpected loud peaks. I've used it that way. But I've never needed that second file so far. Some day it might save my bacon indeed.

But this way my audio comes in pretty similar shot to shot. Time, again ...

Sent from my SM-S908U using Tapatalk

Bob has extraordinary ears and amazing talent

He's also working at a much higher level than I am. For my corporate/event projects, 32 bit Float is akin to shooting RAW video, with results that are perfectly in line with the quality expectations of my clients.

That's not quite true, because there's still effectively a "DMIN" and "DMAX" for camera exposure. You wouldn't (say) deliberately underexpose an image by 5 stops, knowing you could just bring it up 4 stops in post. What happens with noise when you do that? What if part of the image is actually crushed in capture Wouldn't it be better just to expose it correctly to begin with, while still protecting highlights? Learning key-to-fill ratio and how to read scopes is a very basic part of cinematography. And most people shoot with a compensating LUT to bring the image back to normal on set for monitoring.

And that's what good levels are. But it requires an experienced human being to set the exposure -- or in the case of sound, audio gain -- and there is skill and good judgement required. All you wind up doing is taking less time during production to record it way, way, way too low, and then having to fix it all in post. And that puts lots more pressure on the poor editor to even out all those bad levels... something that could have been done during the actual shoot. That's why location sound mixers exist, and that's why DPs exist.

BTW, last comment: one great thing about multitrack location recorders is you can record all the ISO channels -10dB down from normal level, then create a temporary mono (or stereo ) "Mix Track," which is basically used for dailies as a temporary mix. This kind of thing has been done as long as I can remember, certainly at least 30 years, at the very dawn of digital sound. This gave you the best of both worlds: usable sound ready to go on the mix track, with -10dBfs lower-level ISOs for individual wireless mics, separate booms, plant mics, even music playback. And (amazingly), all digital audio software can easily crank tracks up 10dB. That's not that much, relatively speaking.

Anybody who wants more information, read this article on gain staging:

https://www.izotope.com/en/learn/gain-s ... do-it.html

It explains why there are no shortcuts in making sure that sound levels are optimized at each step (mic, mic preamp, transmitter, receiver, recorder, processing, playback). If one of those is wildly low, it's going to have a drastic effect on the end result, both in terms of sound quality and in terms of time used.

BTW, thanks to the people who've commented to me privately on how they, too, are stymied by the belief that you can record at ultra-low levels on set on the mistaken belief that this will avoid problems in post. It actually causes MORE problems. The key is to understand how to avoid clipping in the first place. Note that recording at -50dBfs or even -60dBfs won't help you if the mic preamp is distorted (in other words, analog distortion). You have to very carefully monitor everything as it goes along, and not just trust seeing the meters bounce up and down. Your ears will tell you the truth.

BTW, last comment...

So you just going to continue ignoring the audio comparison between 24bit and 32bit I posted for you and even the null test I explained how to do yourself? What is the point of responding to you if you are just going to ignore the responses?

Anybody who wants more information, read this article on gain staging:

Did you read the article you posted? It says that 32bit can "ameliorate distortions" and "allows for handling audio in ways unthinkable in the days of analog, and impossible in fixed-point digital systems."

If one of those is wildly low, it's going to have a drastic effect on the end result, both in terms of sound quality and in terms of time used.

False. With 32bit it CAN be very low with no problems. Once again I direct you to the null test I mentioned before. Turn down a clip by 100dB and render it out a 32bit float, then load it back in and turn it up by 100dB. It will completely null with the original file. It is identical.

Note that recording at -50dBfs or even -60dBfs won't help you if the mic preamp is distorted (in other words, analog distortion).

Note that 32bit recorders use MULTIPLE MIC PREAMPS AND CONVERTERS internally set at different levels in order to cover a wide range from very soft to very loud sounds without distorting. Instead of you having to ride the gain knob, the 32bit recorder handles this automatically (in other words, NO ANALOG DISTORTION). I've explained this multiple times, provided a diagram and video from Rode and John Paines also provided info from Tascam. But you've ignored all of it.

Here is an even more extreme audio example (which you will probably also ignore). This goes from wiping a cleaning cloth on a car to revving the engine, all without touching the recording device. The clipped engine revving is turned down in post with zero issues. Jump to 1:17.

Tekkerue We are wasting our time. If he now does understands that 32bit float is directed at applications he does not understand. He cannot admit to himself or others he was wrong or not technically competent. So will not respond to this thread in any other way.

Tekkerue We are wasting our time. If he now does understands that 32bit float is directed at applications he does not understand. He cannot admit to himself or others he was wrong or not technically competent. So will not respond to this thread in any other way.

Yeah, I can see that. But even if it just gets him to stop arguing against the feature request, I'll take that too.

I really don't understand the need for certain individuals to argue against feature requests for absolutely no good reason. I've had that happen on my requests too, and of course when I've pushed back they just stop responding so they know they have no good reason to do it. I've been here a little over a year and a half, and I'm already quite sick of seeing it constantly.

My DAW Reaper can boost 150dB at the track level and over 6,000dB at the clip level. If you don't need more than 30dB, great, no one is forcing you to boost more. But for those who do need it, it's there for them. Not a big deal.

Sean, just a couple of observations which are not intended as negative comments against 32bit float.
Your test of turning down a track by 80db, recording it in 32bit float, and then boosting it by 80db is flawed as any low level noise that was in the original recording will still be there having been brought back up to its original level. What do you get noise wise if that signal was originally recorded at say -60db in 32bit float and then boosted up to 0db?
As to the plug in your mics and go scenario, I notice that the Sound Devices MixPres still have input gain controls for setting initial levels. So setting workable levels before recording would still seem to be a sensible move.

Just to clarify what I said about the -60db signal. Let's say that everything I record is around 0db but some of the sounds are at a very low level in that recording that I want to bring up closer to the rest.
If the mic self noise is -114db and I boost that -60db wanted part of the signal up to 0db would the noise now be at -54db?
I don't have access to a 32bit recorder so can't test this for myself.

32 bit float is intended for uncontrolled recording environments NOT a film set with controlled settings. It is wonderful for single operators who cannot manage camera and mic recording at the same time. It is exactly like having one track recorded at a low level and another track at a high level both managed correctly. But rather than then having to edit these two tracks together they are embedded in a 32 bit single track. The editing task is to bring the output back into 16 or 24 bit. Lots of DAW’s can do this correctly RESOLVE Cannot. That is what is being asked for in this request.

I can assure you 32 bit float from F3 works very well. For me it means getting good levels for the actors on stage without clipping for applause or music with no effort on my part at the time of recording.

I understand the value of no clipping with sudden loud bursts of sound, and I understand you getting good levels when recording. The OP however did not get good levels at the time of recording so needs to raise the level a lot. My interest is, in raising the track to a usable level what happens to the noise floor?
For instance the popular Rode NTG video mic has a signal to noise ratio of 79dBA and self noise of 15dBA.

Charles, you can find the noise floor issue, with samples, covered below. What they did was record the same signal in 24 bit and 32 bit float, with the gain settings of the Mix-pre at the minimum -- so much so, you don't even see levels after importing the clips. The results, when gain is applied in post, speak (literally) for themselves.

https://www.sounddevices.com/low-signal-32-bit-float/

Thanks, John, I will have a play.

Having played both examples in Fairlight I can now understand the affect on the noise floor. Interestingly I only had to apply 21db of clip gain to bring the piano note up to a usable level.

Thanks for sharing that link John, that’s a great demonstration of how the gain set on the recorder does not affect the noise floor or reduce sound quality. Of course mic positioning and any noise introduced before the recorder still apply just like they would with 24 bit.

Just to add my voice to this topic (again), lots of documentary and reality TV use 32 bit float recording, so it goes far beyond the solo filmmaker who isn’t monitoring their levels. I work on lots of docu / reality projects in extremely variable environments, and we don’t have the luxury of reshooting a take. What you capture is a once in a lifetime moment that you cannot re-create. We always have a sound recordist and mixer on set who’s adjusting levels for the scene, but sometimes things happen that you cannot anticipate. This is why we choose 32 bit float, because no matter what our levels are set to on the recorder, we know that we are capturing the full dynamic range of the input signal without clipping or introducing unnecessary noise.

I just ran into a scenario where we were capturing extremely loud cymbal crashes, but then in between takes there was a comment that a musician made pretty quietly, totally unanticipated, and that comment ended up making the cut. It was recorded super low on the recorder as we were only expecting to record the really loud cymbal crashes and the comment needed about +45 dB of gain to get it sounding good. Because we were recording in 32 bit float, we were able to increase the gain without any consequence of raising the noise floor beyond what was already present in the mic's self-noise. However, since Resolve’s volume slider only goes up to +30 dB we had to bounce out that individual clip with the gain change to continue increasing the volume. Not the end of the world, but a little annoying. If only our volume slider let us input values beyond +30dB!

We can argue all we want about if 32 bit float is being adopted by Grammy winning engineers or not, but the reality is it exists, people are using it, and I think having a volume slider that takes advantage of some of the insane dynamic range that it offers is a pretty reasonable request. I don’t know why people get so obsessively defensive about a volume slider. Your jobs are safe. A volume slider isn’t going to replace you. We’ll still need sound recordists and mixers on set.

I really don't understand the need for certain individuals to argue against feature requests for absolutely no good reason. I've had that happen on my requests too, and of course when I've pushed back they just stop responding so they know they have no good reason to do it. I've been here a little over a year and a half, and I'm already quite sick of seeing it constantly.

Welcome to the club.

Know that there are people who really appreciate your deep dives into these different subjects. Never change. I personally enjoy your way of explaining things. I wish I had your self control

I looked into this 32-bit float stuff a long time ago. I don't own any recording device, I don't plan to own one in the future (but we never know).

I watched tons of videos about it. Because I'm naturally curious, and it's always interesting information. Many concepts we can learn on very specific subjects can unlock our comprehension of something else. Knowing something knew is never a waste, even if we don't use the information right away.

The problem is that some people live their life based on their emotions.

"I feel like... I don't like it.... I vote  not."

It's a big problem in general because it literally destroys societies when anyone goes against any facts and decisions are made on feelings. They can never give counterarguments based on common sense and logic.

You just need one person to argue against facts, good examples, deep explanations, etc. to ruin a forum.
I left photography forums many years ago because of that that. Always 1 to 3 persons against 10, 30, or 100 users. I'm talking about the era where shooting RAW started to leak into the consumer part of the photography world. The "fight" of those against RAW shooter is dead for 10+ years.

And it's sad when it happens on forums focused on "creative" tools and creativity in general.

It's always about YOU don't need it because I don't need that. Never about YOU should have it because you seem to need that (and I might use it too in the future).

And the same people constantly arguing against good new features will use

"Sooowwwy, take back your 32-bit audio file, the levels seem off, just get it right the first time, I won't use Resolve with its new feature... thank you" - nah, they are going to be very happy that Resolve can deal with this type of stuff.

Same thing with "someone" who was against having Resolve capable of opening MKV files, VP9 and other audio files. I'm sure he'll still transcode everything to stay in line with his old feelings... Right, right? And he will NEVER open these files with Resolve. (MKV is used in serious applications now because it's one of the most flexible container, and free to use! It's everywhere. I've used it for more 15 years myself)

Your test of turning down a track by 80db, recording it in 32bit float, and then boosting it by 80db is flawed as any low level noise that was in the original recording will still be there having been brought back up to its original level.

That is actually the argument I was hoping Marc would make. I was trying to walk him through this piece by piece and getting him to that point was the first part of it. First I wanted to show that his blanket statement of turning up any clip by 120dB will automatically have high noise is incorrect. What matters isn't the amount of boost, it's the signal to noise ratio. If you have a very, very quiet signal, but the noise floor is -100dB below the signal, then you can boost the signal up to a usable level and still have a -100dB noise floor. That is what my example shows.

My next point was regarding the recording process because recording quiet signals with very low noise is exactly how 32bit recorders work by using multiple mic preamps and A/D converters. This means you can boost quiet signals because the noise is much lower than the signal.

In the diagram I posted from Rode, they show four preamps set to different levels so that quieter signals are recorded hotter than loud signals. You'd get the same result by manually riding the preamp gain, except this is automatically done internally. However, by turning up only the quieter parts and not the loud parts, that is reducing the dynamic range (or compressing) the signal during recording. What they'd have to do with their DSP processing is turn the quieter parts back down in order to keep the original dynamic range. To bring this full circle back to my example, they have recorded a quiet signal at a loud level and then turned it back down and stored it in 32bit, then gave that file to you. You can now turn that signal back up in your editor because like you said, "the original recording will still be there having been brought back up to its original level".

There is a method to my madness. This is where I was hoping the discussion would go with Marc, but he refused to engage in the discussion.

As to the plug in your mics and go scenario, I notice that the Sound Devices MixPres still have input gain controls for setting initial levels. So setting workable levels before recording would still seem to be a sensible move.

Absolutely, there are a number of reasons why you might want to set gain for your own convenience in post, realtime monitoring during the recording, etc. but unlike with standard 24bit or 16bit recording, it's not essential for the quality of the recording. I've also seen videos explaining how useful it is in times where you need to capture a spur of the moment event and don't have time to set levels. Not all recordings are done in a controlled environment on a schedule. Since Resolve is trying to be a one-stop solution covering a wide variety of applications, it should be able to handle this too without needing to turn to third party plugins for such a simple task like volume.

Charles, you can find the noise floor issue, with samples, covered below. What they did was record the same signal in 24 bit and 32 bit float, with the gain settings of the Mix-pre at the minimum -- so much so, you don't even see levels after importing the clips. The results, when gain is applied in post, speak (literally) for themselves.
https://www.sounddevices.com/low-signal-32-bit-float/

Thanks for that example. I posted a video comparing 24bit and 32bit recordings earlier, but downloading the files and experiencing it for yourself in your own editor is a much better learning experience.

Welcome to the club.

Know that there are people who really appreciate your deep dives into these different subjects. Never change. I personally enjoy your way of explaining things. I wish I had your self control

Thank you, I appreciate that!

I looked into this 32-bit float stuff a long time ago. I don't own any recording device, I don't plan to own one in the future (but we never know).

Same here. I find audio tech interesting and look into things even if I'm not going to be using it. If I did use it, then I also use Reaper for audio so I have more volume boost than I'd ever need with over 6,000dB of boost at the clip level. I've certainly never needed anywhere near that much, but it doesn't hurt me to have it there. So I do not understand people arguing against it. If you don't need it, then just use what you need. It's really that simple.

You just need one person to argue against facts, good examples, deep explanations, etc. to ruin a forum.

Totally agree! Especially when it's in virtually every single request thread. Then the request gets taken way off topic and useful input regarding how to implement it or discussing problems it might encounter get drowned out by pointless nonsense.

Occasionally I will oppose a certain feature request, but when I do, I have reasons for it and can articulate my reasons. And I'm also open to having my reasons challenged. This is how we can work out the best way to implement a feature or if there is something else that could be done instead but get the same result. The best way to implement the request is the kind of discussion that should be happening in feature request threads. Not pushing back against a small handful of users who too often don't understand even the basics of what is being requested.

Same thing with "someone" who was against having Resolve capable of opening MKV files, VP9 and other audio files. I'm sure he'll still transcode everything to stay in line with his old feelings... Right, right? And he will NEVER open these files with Resolve. (MKV is used in serious applications now because it's one of the most flexible container, and free to use! It's everywhere. I've used it for more 15 years myself)

Yeah, the things people will argue against here never ceases to amaze me. But what they don't seem to be willing to accept is that other people have different workflows than they do and work on different types of projects. The features I need might be very different from the features someone else needs and it's ok if Resolve is flexible enough to work well for everyone.

Reaper is a perfect example of this because its flexibility is usually a top reason people give for using it. It is mind boggling how much you can customize in Reaper and it is extremely light weight, fast, and stable. This goes against the few naysayers that claim having too many options will make the software bloated and slow. Every time I get that response I ask them if they've ever used Reaper, but unsurprisingly I get no response. As a musician I spend far more time in Reaper than in Resolve and it's a frustrating experience when I switch over to work on a video project because of how inflexible and rigid Resolve is. Resolve's tools are amazing and the quality of the final product is amazing, but the process of getting there is not.

Just for those interested. The F3 has a level setting control that alters the headphone output and also how the level is displayed on the LCD. Has no effect on what is recorded. However it is useful to set this well as the visual representation in the LCD while recording is how waveform will appear later while editing. Again has nothing to do really with how the source signal can be manipulated. If it looks clipped it can be brought down, if it is too quiet can be increased. I suppose it is close to shooting RAW for audio. Just like shooting RAW you have to do some things to get close but there after it can be modified in editing.

What happens with noise when you do that?

With 32-bit Float audio? Not much.

Think of it this way, Marc. With RAW video or 32-bit Float audio, you have more room to correct in post.

Not infinite room, just...more. And that's often very helpful when you're doing all the work yourself. (Or even if you're not. )

It's funny when Marc points out Bob Katz...

I do own his book "Mastering Audio the art and the science".
In chapter 5 p.78 Bob Katz is saying about 32-bit float:

In floating point system, you can break all the rules: floating point can literally ignore the individual levels in the chain. It's possible to drop the signal level 100 dB, store the signal as a floating point file, then open the file, raise the gain 100 dB and get back the original signal, with little or no deterioration. Or vice versa, you can raise the signal 100 dB and then lower it and get the same result, providing that intermediate products are stored in floating point format.

But what does 3-time Grammy-winning engineer Bob Katz know...

To get back to the topic. Currently in DaVinci Resolve a clip can only be increased by a maximum of 30dB. The “Normalize” function ultimately only serves the maximum 30 dB level increase. Larger level changes cannot be achieved with this function either. On the other hand, Normalize is a great tool for quick, automatic level changes. I think it would be very good if the "Normalize" function instead placed an "Audio FX" (for example "Preamp") on the clip, with a range of up to 120dB. This would have the advantage that the normal level (max. 30dB) of the clip was not yet used for mixing. The current level path is therefore completely sufficient. So "Normalize" should install a "Preamp". If necessary, you can either use it manually as “Audio FX” or automatically using “Normalize”.

By the way, the idea is not new, that's exactly how it worked with the ancient FCP7.

With the wide range of 32 bit float normalize is likely not the approach to use. Dynamics with makeup is what I use for a simple one shot correction. That way you can bring up the low levels ( with makeup ) to the level you would like and use compressor to control the points that would normal clip or be too loud.

For those that require more than 30db of gain might I suggest the free Blue Cat Audio Gain Suite. I've just downloaded and installed the Win 64bit VST3 version and it opens fine in Resolve. It provides up to +60db of gain so may be just what you are looking for.

https://www.bluecataudio.com/Products/P ... GainSuite/

Blue Cat Audio Gain Suite.jpg (50.22 KiB) Viewed 4171 times

Currently in DaVinci Resolve a clip can only be increased by a maximum of 30dB.

Hence this...

viewtopic.php?f=33&t=177369

Get it right on set

Agreed in general. But remember that 32 bit float recorders (mine at least) don’t have gain controls. There is nothing to get right. It records the full 1,500 db range at all times. You can record the sound of a butterfly and then the sound of a jet engine 10 feet away and they’ll both sound perfect provided of course there is no clipping at the mic. It’s a pretty crazy tech IMO.

(I messed up the quote LOL… this forum is difficult on a phone)

Yeah, make sure you record way too low, so the device self noise now becomes an issue to deal with as well. Which 32 bit won't fix

And yes, 32 bit will need to be delivered in 24 bit to post.

Or, how about you learn proper gain staging and everything will be fine.
No can imagine your surprise when your 32 not recorded audio is noisy and distorted, because you don't understand that your mic pre or mic will distort well before you hit the max level of your 32 bit device.

Oh and btw, 32 bit definitely cause artifacting when hitting max levels. I hear it when dialog is pushed too hard in a 32bit mix path.

And nobody in proffesional audio post uses Reaper.