• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
0

Adobe Audition - Issue with 32-bit float Recording

New Here ,
Jul 30, 2020 Jul 30, 2020

Copy link to clipboard

Copied

I'm having this issue with recording audio in 32-bit where the peaks of the signal that exceeded 0dB are still being cut off when I reduce the gain after recording. I'm not being able to get the headroom that is expected when recording in 32-bit.

 

The reason why I'm having high +0dB peaks is because I'm transferring audio from a Tascam 388 and some of that audio on the reels was recorded a bit in the red. I'm running from the Tape Outs on the 388 and have the input gain at minimum level on the interface that I'm using for transfer (Tascam 16x08) so I can't reduce the signal input volume any further.

 

I'm recording the digital audio into Audition in 32-bit at 48kHz and again, when I try to recover the peaked audio I just get flat, cut off waveforms.

 

Thanks in advance for any replies!

TOPICS
How to , Playback

Views

3.8K

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 31, 2020 Jul 31, 2020

Copy link to clipboard

Copied

You need to attenuate the input signal more. 32-bit FP is very essentially a software fix, and doesn't have any effect on the digitising part of an interface at all; if you overload that, it distorts at this point and there's nothing you can do subsequently to fix it. Your digitiser works as a 24-bit integer device - they all do. So if your input level goes above +8dBv, it's going to distort. What you need is in-line attenuators for the feed into the Tascam. I have to say that most devices with a line input will handle rather more signal than the Tascam does - I would have said that to deal with line level signals from any sort of pro gear, it should really be able to handle +24dBv, not just 8! If you can find attenuators that reduce the signal level going in by 10dB, you should be fine.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
New Here ,
Aug 17, 2020 Aug 17, 2020

Copy link to clipboard

Copied

Thanks for the info, Steve! I purchased some in-line attenuators and that solved the issue. Learning as I go... Thanks again for the help.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Beginner ,
Jul 04, 2021 Jul 04, 2021

Copy link to clipboard

Copied

I'm having the same problem.  I recorded some test footage on an iphone 12 that was intentionally clipped.  Then I imported the footage to adobe premiere pro and used the properties feature to confirm that it IS a 32-bit file and so is the sequence and project.  I then right clicked the audio to open it in adobe audition.  I lowered the volume of the clip and it remained clipped and distorted with a flat wabeform, just lower in volume.  Am I doing something wrong or is there a reason I cannot recover the peaks?  Thanks

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 05, 2021 Jul 05, 2021

Copy link to clipboard

Copied

You haven't got the same problem at all. Your original was overloaded and clipped on purpose. Whatever you do, you can't get over that - this is nothing to do with the bit depth. Distortion resulting from significant clipping cannot be fixed, because there is no accurate reference to what it should be like. The only clipping that can be partially repaired is the odd single clip that isn't too large, and even that repair won't necessarily be accurate. The OP didn't have a distorted original, just an output that couldn't be turned down enough, which is why attenuation fixed it.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Beginner ,
Jul 05, 2021 Jul 05, 2021

Copy link to clipboard

Copied

Thanks for the reply.  I thought 32-bit float can record up to 700 db over 0.  And you can just bring it back down in a daw.  I've seen videos of people doing it.  I have concert footage that I would really like to save but having same problem with that.  

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 05, 2021 Jul 05, 2021

Copy link to clipboard

Copied

You need to understand that recording bit depth has nothing whatsoever to do with distortion. You can't fix distortion. Nothing can fix distortion. And Floating Point processing is only a software trick - real hardware produces integer-based outputs and they have an absolute ceiling of 0dB before they overload. Converting them to Floating point after this (which is the only way you can do it) makes not a shred of difference to that. Floating Point is not a recording format.

 

Yes you can process material, once it's in Floating Point, so that it massively overloads and can subsequently be bought back down without loss. But if that signal is distorted in the first place, then it will be just as distorted as it was when you return it to a more sensible level. So if you overload a mic stage and it distorts, you're stuck with it.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Beginner ,
Jul 05, 2021 Jul 05, 2021

Copy link to clipboard

Copied

Thank you for taking the time to explain this to me.  I do understand that once the audio is distorted, it cannot be fixed.  But I also believe that with 32-bit float, if clipping was the only thing causing the distortion, the volume can be lowered below 0 and recovered.  So I guess I don't understand what actually caused the distortion and how I can avoid it in the future.  Thanks

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Mentor ,
Jul 05, 2021 Jul 05, 2021

Copy link to clipboard

Copied

im pretty sure an iphone can't record true float. some devices record multiple tracks and combine them. the more bits, the more dynamic range you can store. check out some of the new 32 bit recorders. they don't even have a gain knob!

only a low, medium, high for the voltage gain to the preamp for DSLR's.  note: you can still overload the preamp with 32 bit. its still a voltage signal after all.

 

only pure digital information can store full float values correctly.

this is why you can over normalize in some DAW's  and bring them back down again. it's pretty amazing but you have to have a proper pipeline to do it.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 06, 2021 Jul 06, 2021

Copy link to clipboard

Copied

quote

check out some of the new 32 bit recorders. they don't even have a gain knob!

only a low, medium, high for the voltage gain to the preamp for DSLR's. 

 

So what do you think those settings do? That's the gain 'knob'...

quote

you can still overload the preamp with 32 bit. its still a voltage signal after all.

 

only pure digital information can store full float values correctly.

this is why you can over normalize in some DAW's  and bring them back down again. it's pretty amazing but you have to have a proper pipeline to do it.

Please read my other reply; I don't think you've really got a handle on this at all. Preamps have nothing whatsoever to do with 32-bit values. You can turn any digitised signal into a 32-bit value. What the Floating Point system does is to rescale the integer bits so that they are represented by a 23-bit digitised number between 0 and 1, and then there's a sign bit, and eight scaling bits. That's what makes up your 32-bit signal. So it doesn't matter what sort of integer values you have - they can all be stored as Floating Point signals. All you do when you scale the signal is move its range up and down; it doesn't alter the 23-bit number or sign in the slightest, which is why you can't make it clip.

 

32-bit Floating Point digitising is essentially 24-bit audio stored with scaling bits. Unfortunately there are one or two sites around that don't give a clear explanation of what this means as far as recorded signals are concerned. For instance, the Sound Devices site's opening paragraph is well less than clear on the subject; despite their claims, it's still perfectly possible to overload their system front end.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 06, 2021 Jul 06, 2021

Copy link to clipboard

Copied

Floating Point format and clipping are not in the slightest bit related. Any overload causing clipping has to occur before digitisation. Nothing after that makes any difference; it's clipped. Because of the way it works, you can't cause any additional clipping with a Floating Point format - the scaling digits just stop getting any larger.

 

You prevent distortion causing an overload by using a gain control on whatever input device you are using. The whole idea is to stop the analog signal hiting the highest number the A-D converter can output. This is where the concept of  'headroom' comes in; that's the amount of room you leave for any accidental peaks that go higher than your intended maximum. With digital recording, that's normally about 15dB. In other words, the loudest thing you think you'll be recording shouldn't go above -15dB. Depending upon where you look, you might find slight variations in this figure, but I've never heard of anybody using less than 12dB of headroom, certainly for any acoustic-based recordings. I always use -15dB.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Mentor ,
Jul 06, 2021 Jul 06, 2021

Copy link to clipboard

Copied

I am only referencing DSLR's as an interesting footenote of where 32 bit recording doesn't work as anticipated going into a low quality preamp mic input. if you do have a 32 bit recorder like the 

mix pre II, you can record more dynamic range than is normally possible. Please see here for more information.

"The recorded levels may appear to be either very low or very high while recording, but they can easily be scaled after recording by the DAW software with no additional noise or distortion. "

 

https://www.sounddevices.com/32-bit-float-files-explained/

 

if you don't believe me, download the original files here and play with them.

https://www.sounddevices.com/sample-32-bit-float-and-24-bit-fixed-wav-files/

 

I'd also like to add that there is indeed a native maximum spl for mics as well so you'll hit the limits of the mic long before the 32 bit recorder does.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 06, 2021 Jul 06, 2021

Copy link to clipboard

Copied

Like I said, they don't explain it very well, and the last five words are very significant, especially 'additional'. If you have an analogue signal fed into analog electronics (ie, a mic preamp) it will have a performance ceiling at which point it will distort - and if you digitise that, you are digitising distortion. The best input stages you can build are limited by the Laws of Physics to have a dynamic range of about 120dB, and that is easily digitised in its entirety by a 20-bit ADC, never mind a 24-bit or 32-bit Floating Point one. That's not the issue. The issue is that the input is still overloadable by a mic if the gain is set too high - because the mic output from a loud source will, unless the gain is adjusted accordingly, overload the input and cause distortion. There's no such thing as a 'native maximum spl' either - all mics are different and there are many professional ones around that will produce undistorted outputs of levels that your ears could never stand. It's not the limits of the mic you will hit, but the limits of the input stage it's connected to. And that is what Sound Devices don't describe adequately. Their mic preamps are good - I use them - but they aren't perfect and they very much are oveloadable.

 

In a previous existence I spent a lot of time designing mic preamps. I'm reasonably confident that I know a lot more about it than you do...

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Mentor ,
Jul 06, 2021 Jul 06, 2021

Copy link to clipboard

Copied

each mic does indeed have its own listed as rated max spl. you would be hard pressed to find a mic that can withstand >142db (mix-pre's specs)

 

you can theoretically cram more dynamic range into 32 bits,  per spec IEEE-754 supports up to +770 dBFS. Does that make the mix-pre II not a true 32 bit recorder then?

 

I think the idea here is file format. you can store a special kind of data file format("audio signals above 0 dBFS are preserved in the file"), that is compatible with your computer to recover. if you record at 24 bit, it simply clips...

 

If you want to say that 32 bit doesn't exist because the mic and recorder can't fulfill 770 db, then well, I can't convince you otherwise.

 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jul 07, 2021 Jul 07, 2021

Copy link to clipboard

Copied

LATEST

A Sennheiser MD 421 can easily cope with way more than 142dB. 150dB with effectively no distortion when Sennheiser tested it, and they say that there is effectively no upper limit. And it's by no means the only one.

 

And I don't know where you got the idea that a 32-bit FP system has a dynamic range of 770dB because it doesn't - it's 1528dB.

 

And as for the idea of you convincing me of anything, especially by trying to put words I didn't use into my mouth... well, let's just say that it ain't gonna happen. In just the same way that I can't convince you of why, I expect.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines