Float 32 editing

Which of you "experts" here, works fulltime mixing high end audiopost, everyday?

Note to all: Mark Hensley is an award-winning re-recording mixer & editor in Vancouver and Los Angeles, nominated for three Emmys (won one), and is one of the most talented sound people I know. He's a rare "sound guy" who also understands cinematography and color, a genuine Renaissance man.

Go check out his IMDB credits if you like. Mr. Hensley not only talks the talk, but he also walks the walk.

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.

Are you talking integer or float? Those are computationally different things.
Can’t see how you can push true float and get it artifacting…

But one single tool in a 32bit float path not working in float can break the float computation.
The same like when you get a non-32bit aware plugin on an EXR image file which will truncate the super whites above 1.0

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.

Are you talking integer or float? Those are computationally different things.
Can’t see how you can push true float and get it artifacting…

But one single tool in a 32bit float path not working in float can break the float computation.
The same like when you get a non-32bit aware plugin on an EXR image file which will truncate the super whites above 1.0

Float.
And yes, I hear it.
But then I'm the guy who's been complaining for years about the awful boxy artifacting of Izotope RX broadband reduction, where everyone else drank the kool-aid. And now suddenly people are saying "yeah, RX isn't so great anymore".

It never was.

Float.
And yes, I hear it.

When you can hear it, then you can quantify it by showing that the waveform got altered.
Push the audio in 32bit float until you can hear it, than revert the levels. Now phase subtract the new waveform from the original. If it got butchered there should now be artifacts remaining and be visible above/below the zero line.

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.

Then bring the levels down. The only way you should hear artifacts/clipping is if the levels are too high on the timeline ( your monitor mix is overloading not the clip ) or the mic was too close and it was the mic overloading ( unlikely for dialogue ) You cannot mix 32bit float like 16 or 24bit rate audio. The waveform from a 32bit float recorder is a created waveform not an indication of audio level as that will never be too low or too high. Certainly on the F3 I set mine so that I can see something is being recorded and will look reasonable on the timeline it is just for display and nothing to do with recording levels. I know this is a difficult concept when moving to 32bit float since what you see as a waveform is not what you can get finally. It took me a while to use the F3 as at first I did not set this recording level display correctly and either had a waveform that was close to invisible on the timeline or looked like it was total clipping. However in both cases got the audio I wanted. Now set so that the LCD display looks like something is recording half filling the display and that has been fine.

Two points:

1. The other forum thread you guys mention is not a request, and has no mention to the 30dB limit

2. Most of you guys consider dialogs and its headroom on this thread, but you miss working on really subtle sounds beyond common dialogs.

On my case I'm working on ASMR like videos where my concern is to bring those -60dB to -10dB levels. For this I use Fabfilter L2, but sadly the clip waveform graphs on DVR show no waveform on such levels when increased to 30dB.

I really need to increase levels up at least up to 60dBs to 120dBs to bring to life such subtle audios on the pre-plugin or pre-dynamics stage to be able to use clip waveforms on Fairlight.

Two points:

1. The other forum thread you guys mention is not a request, and has no mention to the 30dB limit

2. Most of you guys consider dialogs and its headroom on this thread, but you miss working on really subtle sounds beyond common dialogs.

On my case I'm working on ASMR like videos where my concern is to bring those -60dB to -10dB levels. For this I use Fabfilter L2, but sadly the clip waveform graphs on DVR show no waveform on such levels when increased to 30dB.

I really need to increase levels up at least up to 60dBs to 120dBs to bring to life such subtle audios on the pre-plugin or pre-dynamics stage to be able to use clip waveforms on Fairlight.

This thread is specifically about dialog. Not some irrelevant subtle audio that has nothing to do with dialog. Sound that subtle on set will be obliterated u ambient noise, and replaced in post.

This thread is specifically about dialog. Not some irrelevant subtle audio that has nothing to do with dialog.

Are you sure? Here is the original post:

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.

This thread is specifically about dialog.

NO! Unfortunately, the topic has taken on a life of its own.
DaVinci Resolve is capable of importing, editing and exporting 32-bit audio.
The creator's request was whether the +30dB limit (without further cascades) on the volume control could be removed. To make it easier to edit the possible higher dynamics of 32bit audio. Nothing else.

I work for television, with a focus on reports. I also like to have a capable sound assistant with me during filming. But it's not a defined film set, but real everyday life. I don't want to waste my time in post-production with 32bit recording. I need a proper mix on site. But I'm happy about the technical possibility of having a 32bit backup. In case the sound engineer is overloaded with several lavaliers and an additional boom, everything is already mixed on two channels. This usually works and that's a good thing. But having separate channels as an option, completely independent of the operator's mix, can help. Even if it means more work in post-production, I can repair problematic situations. But I'd be happy to forego it. Please don't misunderstand this as "we fix it in post". I find that an impossible attitude.

The currently maximum possible increase is actually enough for me. But there doesn't have to be a limit. I can understand the request.

Dialogue or not, if the fundamental claim about the noise floor with 32 bit float is wrong, I think most people would want to know.

We've had samples linked to above -- recordings at very low levels, way below what anyone would do intentionally -- which appear to demonstrate the virtues. Are samples available which demonstrate the "artifacts"?

... but he won an award! (which means nothing honestly).
This FR is going to be implemented. And I bet on Davinci Resove 19 (or a 18.8 version).
If not in one of these versions, it WILL be implemented in the program. It's just a matter of time.

The only positive thing with hijacked thread like this one is all the information coming out from people who know their stuff, like Tekkerue and others who spent time to explain what is this 32-bit float audio thing.

... but he won an award! (which means nothing honestly).

And has worked fulltime proffesionally as an audio engineer for 30 years. Which is probably longer than most who are posting here have been alive


So yeah, there's that

... but he won an award! (which means nothing honestly).

And has worked fulltime proffesionally as an audio engineer for 30 years. Which is probably longer than most who are posting here have been alive


So yeah, there's that

For 30 years, but didn’t yet manage to suppress a slightly arrogant undertone…
BTW, I’ll turn 50 next year, have a mechanical engineering degree, have a production company running for over 16 year and have done the audio mastering for several feature films in my country - if that’s now the way how we counter argue in technical discussions…

I have no doubt you are a respected and talented audio engineer, but why do you and Marc still bring strawmen arguments against 32bit float recording?

The feature I mastered last year was fully 32bit float (Sounddevices MixPre10 II and Schoeps Collette mic sets) from start to finish and the sound department on set had no problem with that, neither did we in post. Could we have done that with 24bit? Sure, but it was fun to explore a new tool and a different workflow.

The idea that it takes 30 years to understand the meaning and practical implications of 32bit float vs 24 int recording is quite silly. You can teach this to a student in a day and he'd acquire the same knowledge and tell you that Resolve indeed needs tools to manage this format. It's not some artistic vision you acquire over the years, it's just simple learnable facts.

The idea that it takes 30 years to understand the meaning and practical implications of 32bit float vs 24 int recording is quite silly. You can teach this to a student in a day and he'd acquire the same knowledge and tell you that Resolve indeed needs tools to manage this format. It's not some artistic vision you acquire over the years, it's just simple learnable facts.

Come back when you're working with extensive sound teams on tight deadlines on A-List ost productiins.

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

Have you ever used any 32bit recorders or done controlled tests to compare 32bit and 24bit recordings? Can you produce test results that contradict the audio examples that have already posted in this thread? If you can, I'd be interested in hearing them.

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.

The max level of 32bit is +770dBFS, so 0dBFS is nowhere near the max level of 32bit. A full mix path is not a controlled test to isolate the impact of 32bit. Doing null tests using volume only without any plugin or hardware effects and then rendering out 32bit float point wav files instead of playing through a 24bit DAC is a very simple way to test this. This will eliminate all other variables that could potentially cause artifacts when pushed to max levels.

I took a vocal clip, boosted it so that it peaked over +95dBFS, rendered it out as a 32bit float point wav, loaded it back in, decreased it back to its original volume, inverted the polarity and it nulled with the original file showing -INF on the meters. Have you gotten a different result from this? Or can you hear differences in two files that null?

And has worked fulltime proffesionally as an audio engineer for 30 years. Which is probably longer than most who are posting here have been alive. So yeah, there's that

But you have yet to respond to any of the explanations, diagrams from manufacturers, or audio examples that have been provided here regarding how 32bit recorders are different from 16bit/24bit recorders. As an analogy, just because you are a full time professional award winning race car driver that doesn't make you an expert auto mechanic. They are different fields.

It's not some artistic vision you acquire over the years, it's just simple learnable facts.

Yep, and it's testable. The noise floor can be measured and null tests can confirm or deny the existence of artifacts. While there are plenty of areas of sound/music that are subjective, this really isn't one of them.

Having just done a test with my Zoom F6, I still see no advantage for recording 32bit at - 40
It does indeed sound fine.
However.

The dialog editor will have to start with processing all audio files up by about 60db to just see the waveforms , as protools only does 36db of clipgain.

This is simply not an efficient workflow.
When production learns you're causing issues down the line, you can kiss your job goodbye.

... but he won an award! (which means nothing honestly).

And has worked fulltime proffesionally as an audio engineer for 30 years. Which is probably longer than most who are posting here have been alive


So yeah, there's that

That's great, but I don't know who you are. For me, you are just someone on a forum who doesn't seem to understand what people are saying.

I'm only judging what you are presenting vs what other are saying, nothing else.

You were introduced as an "award winner", which is fine, but honestly (and respectfully), who cares? This means nothing in this context. It doesn't matter if you have an entry on imdb or anywhere else.

Extremely knowledgeable people here explained how it works (and in other threads about the same subject), why it matters, how it can be used and how practical it is.

I'm far from being a specialist, but I know what's RAW is (for an image), and 32-bit float seems to be the equivalent for audio. Both are only limited by the device capturing the information. If a sensor could capture 10000 stops of light, you would have 10000 stops to work with.

Does it mean that you shouldn't care about how you expose your scene, disregard every basic rules? no. Can you be sloppy (or make mistakes) and still get the same results as if you got it right in the camera, not having days of work ruined? yes.

It's like an MD who doesn't understand nutrition (most of them don't, and don't try to educate themselves on the subject), then starts to argue with those who are not doctors but are very well educated on the subject, by using their "Doctor" title to make it look like they can't be wrong.

Being a full-time professional at anything is not an indication that someone is good at their job either (it's just a general observation and not aimed at anyone).

I respect A LOT those who have a ton of experience in their craft and stay humble. While constantly wanting about what's new in their field.

I won't hijack this thread any further, so +10 for the FR.

I'm far from being a specialist, but I know what's RAW is (for an image), and 32-bit float seems to be the equivalent for audio. Both are only limited by the device capturing the information. If a sensor could capture 10000 stops of light, you would have 10000 stops to work with.

Recording at -40db because you're using 32bit float, is the same as shooting braw at 4000 iso, underexposing to the left, while thinking "not my problem. At least it's not overexposed. They can fix it on color"

Having just done a test with my Zoom F6, I still see no advantage for recording 32bit at - 40
It does indeed sound fine.

Thank you for doing the tests and confirming that it sounds fine.

In a studio environment or fully staffed professional movie set, there very well may not be any advantage. But for small or solo indy creators, this could be a lifesaver. While looking for audio examples to provide for Marc earlier (I don't have a 32bit recorder myself, I've just looked into it) I found an example of someone who recorded wiping a cleaning cloth on his car and then revving the engine in the same shot without having to touch the recording device. The cloth was set to an audible level and engine revving was heavily clipped and distorted. Then he brought down the level in post with no problems and the engine revving sounded totally clean with no clipping.

As a solo creator who is doing everything themself, which option requires less work:
1) Stop filming and reset the gain level for each scenario (what if you wanted both in the same continuous take?)
2) Re-shoot due to clipped audio
3) Restore it in post with restoration software
4) Press the normalize shortcut in your DAW to bring it to a usable level

I know I'd take number 4 any day. And this is even a task that could be scripted.

Not everyone has the same needs and workflow. If you don't need a particular feature and you can't find any use for it in your workflow, that's fine, no one is forcing you to use it. I've never needed anywhere close to the over 6,000dB boost that Reaper provides, but it doesn't hurt me to have it there. In fact, Reaper even has an auto-mute feature to keep you from blowing out your speakers in the event that you accidentally add 99dB of boost when you were only trying to add 9dB. They've thought of everything.

The dialog editor will have to start with processing all audio files up by about 60db to just see the waveforms , as protools only does 36db of clipgain.

It's quite common for there to be hiccups when new technology comes out and a bit of a waiting period for other technology to adapt to it. But that is no reason to avoid moving towards it. If ProTools doesn't allow more than 36dB of clip gain, then that's an issue to take up with Avi... errr... Symphony Technology Group. That's exactly what we are doing here with BMD.

As a solo creator who is doing everything themself,

That's exactly what I am.
I light, direct, shoot, edit, do sound, color. I still don't record my audio at -40.
Your recording devices input/mic pre and Lav microphones are going to distort by the time you hit 144db.

As far as the recording of wiping a cleaning cloth on his car and then revving his engine, my experience in audio post tells me that the wiping sound will be useless due to the sound of ambient broadband noise, and will be replaced by foley. So thats a nice anecdote, but in the real world it's pretty meaningless.

So yeah, it's all fine for one man band, in house doing everything yourself. But when you get into the real world of actual real production with teams on a tight deadline, it's a whole other ball of wax.

As a solo creator who is doing everything themself,

That's exactly what I am.
I light, direct, shoot, edit, do sound, color.

What I mean is entirely solo, as in they are also the talent on screen as well as doing all of the production, which means, setting and forgetting... or lots of stopping and starting to readjust gain levels. There are plenty of solo content creators doing everything by themselves and anything that will provide a wider buffer to avoid having to stop and start or re-shoot I'm sure is welcome.

As far as the recording of wiping a cleaning cloth on his car and then revving his engine, my experience in audio post tells me that the wiping sound will be useless due to the sound of ambient broadband noise, and will be replaced by foley. So thats a nice anecdote, but in the real world it's pretty meaningless.

Unless the natural ambient noise is the sound you want. He was filming outside, so yes, there was natural ambient background noise. It was also a video about 32bit recorders and the example was extreme to demonstrate the wide dynamic range. Restoring the clipped audio was part of the demonstration.

There are people here on the forum using 32bit recorders for their projects and also receiving 32bit files from clients. The latter is yet another reason for Resolve to remove these unnecessary limitations, if you receive projects from clients then you will have no input over how they recorded it. Resolve should be flexible enough to handle these projects too as that will make your job easier when working with clients.

Resolve already has lots of restoration tools like noise reduction, image stabilization, relight, voice isolation, etc. for any one of these features you could make the same argument of "just record/film it properly", yet I don't see anyone having a problem with these tools being in Resolve. I see no reason why simple volume adjustment should be treated any differently, especially when it comes at no negative impact to the final product. IMO this is a no-brainer.

What I mean is entirely solo, as in they are also the talent on screen as well as doing all of the production, which means, setting and forgetting... or lots of stopping and starting to readjust gain levels. There are plenty of solo content creators doing everything by themselves and anything that will provide a wider buffer to avoid having to stop and start or re-shoot I'm sure is welcome.

.

Bedroom hobbyists making YouTube videos.
Recording with 24bit is the least of their problems.

Bedroom hobbyists making YouTube videos.
Recording with 24bit is the least of their problems.

Maybe so. But like I mentioned, Resolve already has plenty of restoration features that will help with other problem areas, including literally relighting the shot. Providing enough volume boost to work with 32bit recorders isn't any different, it's just a feature. For those that need it, they'll be glad to have it. For those that don't, then it won't hurt anyone by having it. Not a big deal.

Recording at -40db because you're using 32bit float, is the same as shooting braw at 4000 iso, underexposing to the left, while thinking "not my problem. At least it's not overexposed. They can fix it on color"

That is a very apt analogy.

I'm reminded of a film student who excitedly told me years ago, "hey! Now we have 8K cameras! That means all we have to do is shoot a wide shot at 8K, and then blow up the image so we have a medium shot, a 2-shot, and close-ups!"

Recording a much too low a level and then having to crank it WAY UP in post is self-defeating. Again, people have to consider the noise of the mic preamps and the microphone itself. All I'm saying is: record the dialogue on set about -10dBfs or -12dBfs low, so that if somebody screams at the top of their lungs, the signal will never clip. It doesn't take a lot of skill to do that. Hell, you could record the levels -20dBfs low and still have more than enough range to normalize them. And again: dialogue changes on every single take an angle, because it's the nature of acting: you need a human being to monitor the levels and make good decisions at how to keep them at a "reasonable," consistent level.

Last comment: I spoke to a good industry friend of mine at Warner Bros. yesterday and he asked me what was going on lately. I mentioned this thread, described the o.p.'s need to turn an audio signal up more than 30dB, and before I could continue, he interrupted and said, "jesus, if I was working on a project where somebody was recording audio so low we had to crank it up +30dB or +40dB, I'd do everything I could to get that person fired! That's wrong for a dozen different reasons!" And note this is a video guy, with 20 years' experience in post-production. He got the problem immediately. He also commented: "I bet this came from a student filmmaker." I have no idea, but you wonder. I have a couple of hundred IMDB credits, and I know Mark Hensley does, but the rest asking the questions...

[But like I mentioned, Resolve already has plenty of restoration features that will help with other problem areas, including literally relighting the shot.

That is not true: we depend heavily on the cinematographer to get the lighting right on the day. All a colorist should need to do is bring out the best in the lighting that's already there in the scene. "Relighting" is kind of a misnomer: much of what we do is placing shadows and adding emphasis to certain parts of the shot. The OFX Relighting tool is more limited than you might think, and it by no means replaces the work of the DP.

We now return you to our audio level discussion thread.

Recording a much too low a level and then having to crank it WAY UP in post is self-defeating. Again, people have to consider the noise of the mic preamps and the microphone itself. All I'm saying is: record the dialogue on set about -10dBfs or -12dBfs low, so that if somebody screams at the top of their lungs, the signal will never clip. It doesn't take a lot of skill to do that. Hell, you could record the levels -20dBfs low and still have more than enough range to normalize them. And again: dialogue changes on every single take an angle, because it's the nature of acting: you need a human being to monitor the levels and make good decisions at how to keep them at a "reasonable," consistent level.

Last comment: I spoke to a good industry friend of mine at Warner Bros. yesterday and he asked me what was going on lately. I mentioned this thread, described the o.p.'s need to turn an audio signal up more than 30dB, and before I could continue, he interrupted and said, "jesus, if I was working on a project where somebody was recording audio so low we had to crank it up +30dB or +40dB, I'd do everything I could to get that person fired! That's wrong for a dozen different reasons!" And note this is a video guy, with 20 years' experience in post-production. He got the problem immediately. He also commented: "I bet this came from a student filmmaker." I have no idea, but you wonder. I have a couple of hundred IMDB credits, and I know Mark Hensley does, but the rest asking the questions...

Jesus, how can someone be so stubborn - after all those explanations, disclaimers and provided examples still beating the old drum like having learned nothing from those long threads...

Recording a much too low a level and then having to crank it WAY UP in post is self-defeating. Again, people have to consider the noise of the mic preamps and the microphone itself.

Did you miss the part where Mark said he just did a test with his Zoom F6 and "It does indeed sound fine"?

Last comment: I spoke to a good industry friend of mine at Warner Bros...

Why was the mention of 32bit recording omitted from your discussion?

But like I mentioned, Resolve already has plenty of restoration features that will help with other problem areas, including literally relighting the shot.

The OFX Relighting tool is more limited than you might think, and it by no means replaces the work of the DP.

I never said it was a full replacement, I said it will help. Likewise, the voice isolation is also not a full replacement for recording voice without any background noise, but it will help. Like I said before, you can apply this to any of the restoration tools in Resolve but were you arguing against Resolve getting these restoration tools as well?

However, when it comes to 32bit recorders, there is no negative impact on the sound quality by recording low and boosting in post. The only actual arguments that can be made against it are that of personal preference and convenience, which are subjective. Also, what is convenient can change depending on the situation. More options means people can choose what best suits their workflow and their needs for the given situation. If it doesn't fit your workflow or is more inconvenient for you, then don't use it.

Recording at -40db because you're using 32bit float, is the same as shooting braw at 4000 iso, underexposing to the left, while thinking "not my problem. At least it's not overexposed. They can fix it on color"

That is a very apt analogy.

I'm reminded of a film student who excitedly told me years ago, "hey! Now we have 8K cameras! That means all we have to do is shoot a wide shot at 8K, and then blow up the image so we have a medium shot, a 2-shot, and close-ups!"

Recording a much too low a level and then having to crank it WAY UP in post is self-defeating. Again, people have to consider the noise of the mic preamps and the microphone itself.

You both still after all these efforts don't understand what is happening with 32bit recorders. The pre-amp is not a singular thing. Think of it as recording with a sensor 'native' at 4000, but also one at 100, 200, 400 etc... They all overlap and they all combine their usable data that is above their respective noise floors, similar to bracketing for HDR photograhpy. While you could add a 'gain' scalar to the hardware there doesn't need to be a recording level, there is only the physical level of what is being recorded. If that what you are recording is close to the surround physical noise of course you have a bad recording. But the file container or the pre-amp have nothing to do with that. How this is scaled in the file practically doesn't matter. After normalization it would still be the same signal because the data range spans much further than needed for non destructive reconstruction.

Yes both appear like technicians who were taught some while ago and have not updated their education so carry on doing the same old things. Before they tell me I am too young to understand I am 81 years old will be 82 in January so likely older than either of them.

To Marc and his comment on the young kid and 8K. I record with 6K and 4K cameras and do indeed crop in, pan , zoom in a 1920x1080 project. Works great. Just like using my F3 for audio. When you record events by yourself, with only one shot at getting the video or sound every technological help is of value. If you are on a set with a crew and can do a re-take fine , but that translates to money so still not a great option.


The fact that Marc and Henchman seem incapable of understanding the technology does not give them the freedom to critique others who are more up to date with their education.

The fact that Marc and Henchman seem incapable of understanding the technology does not give them the freedom to critique others who are more up to date with their education.

To be fair, Henchman actually did come back after doing some testing with his Zoom F6 and said:
"It does indeed sound fine."

After that, all he said was he still didn't see the need for it, which is entirely subjective. If he doesn't see the need for it in his workflow, then I'm not going to argue with him that he has to use it himself. My main concern was debunking the false statements regarding the technical side of 32bit recording like high noise, artifacts, etc.

If you want to put more effort in during recording to get the levels you want in post. That's fine.
If you want to record low and then put in more effort to adjust levels in post. That's also fine.

There is no negative impact on the sound quality so neither way is right nor wrong, it's simply whatever works best for you. If you have a preference, then do it that way. Resolve removing the arbitrary limit of 30dB boost is going to be very useful for those who would prefer to record low and then boost in post.

You are correct in that he at least did a test. The confusing bit there is he says he recorded at -40db. I am not sure how you do that with a 32bit float recorder since it really ignores any input settings other than settings for display or headphone. This does have a very small effect on noise level. I think tests I have seen with the lowest settings giving noise floor of -78db and the highest setting at about -82 db for the Zoom recorders. Also believe you have to set this headphone and display setting before recording. After recording starts the headphone level can be adjusted but the display level is fixed and what ever effect this has on recorded file.

I also think everyone can choose what they do. Just don't criticize someone who has a different point of view and need for their application.

Also agree Resolve should correctly support 32 bit float audio. Not really a problem in my application but would be nice to know that full range of adjustments is there if needed.

A simple If 32-bit float, enable expanded sliders would do.


Sent from my iPhone using Tapatalk

A simple If 32-bit float, enable expanded sliders would do.

I don't think checking the file type is necessary. Assuming they have the max volume stored in a constant, just change it from 30 to something ridiculously large. My DAW Reaper doesn't care what the file type is and you can boost clip gain and track volume over +6,000dB.

I would like to thank all of you who have explained the technology to give a better understanding of 32bit Float. That's now locked into the grey matter.
In defence of myself, Marc, and Mark, I would say that as we have spent our working lives in controlled environments, ie studios, where the use of 32bit Float didn't occur, it passed under the radar.
Re Reaper, I have to agree with Mark.

If anyone wants even further confirmation, Sound Devices published a test of an instrumental recording with MixPre levels set "accurately", set 40db too low and set 30db too high. All three recordings were identical:

https://www.sounddevices.com/noise-in-32-bit-float/

And for the mathematicians here, the numbers are explained:

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

Excellent John. On this and the F6 it is of course possible to move gain controls as indicated but in fact this does nothing for the actual recording other than the waveform display. On the F3 it is the same for indications on the LCD and the headphone output. Also of interest in that they use Reaper.

I would like to thank all of you who have explained the technology to give a better understanding of 32bit Float. That's now locked into the grey matter.

You're welcome.

In defence of myself, Marc, and Mark, I would say that as we have spent our working lives in controlled environments, ie studios, where the use of 32bit Float didn't occur, it passed under the radar.

I don't fault people for being unaware of 32bit recording and how it works. I've never used it either, I'm just interested in audio tech and have looked into it. Likewise, I don't use ProTools (I've only used it a small handful of times and I didn't care for it) but I somewhat keep up with new developments out of morbid curiosity.

I do fault stubbornly digging your heals in and being unwilling to listen to new information. You never did that though. Mark initially started out that way, but eventually came around... and well, still waiting on Marc. Maybe he'll come around eventually too.

Re Reaper, I have to agree with Mark.

Sure, I understand that ProTools still has a stranglehold on the industry. But I mentioned Reaper because it has already implemented the feature we are asking for in Resolve. Reaper has more boost than I've ever needed and it hasn't hurt me at all by having it there. In fact, Reaper has an auto mute feature to guarantee you don't hurt anything (like your ears, speakers, etc) if you accidentally add more gain than you intended. Maybe there are other DAWs that have this too, but I'm not aware of any.

Just looked this up. Pro Tools has Noise Burst Peak Protection which was introduced in 2019. It will auto mute a track where the volume suddenly exceeds 20dbfs providing protection when using 32bit float audio.
I don't have this on my PT system as I froze it at version 2018.4 when I retired from doing commercial work.

Just looked this up. Pro Tools has Noise Burst Peak Protection which was introduced in 2019. It will auto mute a track where the volume suddenly exceeds 20dbfs providing protection when using 32bit float audio.

Ah ok, so there really isn't any reason for Pro Tools to not remove the arbitrary 36dB boost limit.

Resolve might want to include a similar auto-mute protection when allowing for large gain boosts just to be on the safe side. Typos happen.