Why 24 bit audio and anything over 48k is not only worthless, but bad for music.

[gregorio]

Your first response was to me saying "32bit workflow" doesn't exist.

Which is correct, a 32bit workflow does not exist. A 32bit recorder is not a 32bit workflow.

When I asked to first look at another article from ProTools editors about one of the new 32bit devices, it's more weeding through insults and off topic content like US gun laws

Then you’re contradicting yourself! I mentioned gun laws to refute an example usage scenario/application in that article YOU asked me to look at and You specifically asked about applications.

I'm giving up, it seems you can't read.

Right back at you, even of what you yourself have asked?!

I didn't insinuate that a recording goes to the full 32bit: only that it has plenty of "headroom" for any recording situation.

Yes, it does have plenty of headroom and so does 24bit. Except, *possibly* for the recording situation that never existed previously and that they themselves have created, of having no control over the gain. And, except for reducing workload on the production recordists (and passing it on to the next poor sap in the workflow chain), there appears to be more disadvantages than advantages to having no control over the gain, though this is unclear currently.

The question was why do these devices exist if a ADC goes to 20bit. Now there could be useful info about how that's accomplished (if using 2 ADCs) and processed.

We could have a discussion about why these devices exist but I’m not sure, given the above, that I wouldn’t be wasting my time. We couldn’t have an informed discussion about how they’re accomplishing using 2 ADCs per channel though, because as far as I can see none of them have discussed it publicly and I can understand why. There maybe something in the patent applications but I would think even there it’s obscured.

G

 

[Davesrose]

Right back at you, even of what you yourself have asked?!

G

I'm sorry, I just remembered you brought up US gun laws. Maybe I should have gone back to the post and reference any other paragraph before actually referencing the article.

It seems "32bit workflow" has been uttered to distinguish between the new 32bit recording devices (that there are audio forums with ProTools using this term). I understand your point that it doesn't refer to the whole workflow, that has different bitdepths with editing and delivery (so you would not use it: future reference I'll try to make a distinction with recording device).

Now here is the part about cameras that you didn't read: a camera's recorded bitdepth is 16bit (even though with any situation and exposure settings, the ADC won't fill that whole dynamic range). Perhaps I was making it more complicated by saying we can process multiple exposures for one 32bit file when editing. But I included it as an example for why it could be in the editing stage.

 

[VNandor]

Am I already using a "non-24bit" workflow if I'm recording straight into a DAW? Obviously if I'm overdriving the mic preamp it doesn't matter if I turn down the digital signal, the sound is still clipped. Or is the clipping from the ADC if I set the gain too high? If only the digital signal is clipped during the recording I can turn that down and the clipping is gone for me.

 

[gregorio]

I'm sorry, I just remembered you brought up US gun laws.

No problem. This type of exchange is common when we argue with audiophiles but in this case it’s refreshing to get an apology. “Unexpected” events suggested as the main usage scenario for 32bit in the article, during live music performance and filming *could* be scenarios where there *might* be a benefit to 32bit recording. But the actual examples given (in one case, of cowboys whispering and then shooting a gun) would not be unexpected, in fact it would be illegal if it were. The live music example of the musicians changing their levels/settings would only be unexpected to a sound engineer who’d never done the job before and the aircraft example would be a problem regardless of 32bit.

It seems "32bit workflow" has been uttered to distinguish between the new 32bit recording devices (that there are audio forums with ProTools using this term).

Yes, because at that point in the process (the use of ProTools in audio post) we actually have a “workflow” or at least a significant part of one. Once the production sound team have finished, they’ll turn over all the production sound recordings to the Picture Editor. Picture Editors generally only have a rudimentary understanding of audio. Their NLE will probably convert the recordings to 24bit on import and many/most of the files will therefore appear faulty to the editor, with no way of fixing them. If the editor has been taught how to maintain the 32bit files and how to deal with them, they will then have to apply gain-staging (and still maintain the 32bit files) and almost certainly will make some/numerous errors when doing this. When they’re done editing they’ll have to export all the audio in their timeline to the audio post team. By far the most common way of doing this is with a container file format called an AAF but AAF only supports up to 24bit files. There are potential ways around this (referenced instead of embedded audio files) but a 30min job for the editor and audio post team could become a several day job, plus, all the gain-staging will have to be done yet again (on 10,000 or so audio files!). So, now we really are talking about workflows and depending on the existing workflow, the gain-staging and maintaining 32bit will, in the majority of cases, at least be problematic and in some/many cases unworkable.

Now here is the part about cameras that you didn't read: a camera's recorded bitdepth is 16bit (even though with any situation and exposure settings, the ADC won't fill that whole dynamic range).

That’s not THE part about cameras I didn’t read, it’s just one of several parts about cameras I didn’t read! Camera/digital images is off-topic, I have little interest in them and besides a relatively small amount I need to know to do my job, I know next to nothing about them and so can’t make any informed comments or procedural comparisons.

I have no idea what bit depth cameras actually work at, all I can talk about is audio. The vast amount of location sound recordings I deal with have a dynamic range of around 30dB or lower (3-5bit). Most of the Iso (mic) tracks are wireless lavs, which have a practical dynamic range of about 40dB or so (about 7bit equivalent). 60dB of dynamic range (10bit) is very rare and the actual dynamic range of a top mic like the CMIT 5 (81dB) is never achieved in practice. So in most cases 16bit, even with very significant amounts of headroom, is serious overkill and 24bit hugely ridiculous overkill. 32bit float is literally many quadrillions times of beyond hugely ridiculous, although without any gain control there *might* be some very rare/extreme scenarios where 24bit is *possibly* not quite enough. Having no gain control is a serious hurdle to overcome, are you aware of how much is necessary?

Am I already using a "non-24bit" workflow if I'm recording straight into a DAW?

TBH, the whole thing is pretty arbitrary if we look at the actual details. The actual A to D conversion is probably 5bit to 8bit (with a 15MHz or higher sample rate), then it’s decimated down to 32 or 24bit passed to your computer and then, depending on the architecture of your DAW, 64bit and truncated to 24bit.

Obviously if I'm overdriving the mic preamp it doesn't matter if I turn down the digital signal, the sound is still clipped. Or is the clipping from the ADC if I set the gain too high?

That depends on the level of the acoustic signal, the output level of the mic and the characteristics of the mic-pre. If you’re overdriving the mic-pre or the mic itself, turning down the digital signal makes no difference, regardless of bit depth. Only if it’s digital clipping in the ADC during/after decimation will 32bit make a difference. In that case only; lowering the digital signal (below 0dBFS) will result in a proportionally lower signal without the digital clipping but with 24bit you’ll get a proportionally lower signal with proportionally lower clipping.

G

 

[71 dB]

There are now people who are using 32bit workflows. I asked why?

They are using 32 bit floating point recording to be exact because it offers the benefit of not having to worry about gain staging while recording. It is not to record better sound quality (that's limited by the mics). It is to make the job easier and "fool proof" to yourself (and potentially harder for other people, but who cares, right?)

... insisted it was only applied with mixing and not recording. I have supplied an example of a recording device that records a 32bit file.

Yeah, but the 32 bit file is used to remove the need for gain staging while recording.

 

[Davesrose]

Here we go again. We do NOT have sound files that “go to the full 32bit”, in fact it’s not even remotely possible in practice or theory. The loudest sound that can exist is at 194dBSPL, the quietest place on earth has a noise floor of about -20dBSPL, so a dynamic range of 214dB covers everything possible on this planet. If 32bit float had a dynamic range of 454dB that would be a trillion times more than anything possible but it doesn’t, it has 1,536dB dynamic range! I don’t see any correlation between digital images and digital audio if you can fill 32bit float.

G

Here is you not understanding a camera records at 16bit, and due to exposure settings, doesn’t fill up that entire range! You can’t read one paragraph from me, and you expect me to remember verbatim a previous post that goes half a page. Harder still that you break up a person’s post to lose their context while you talk down to them (and later admitting you didn’t first read their post).

 

[Davesrose]

They are using 32 bit floating point recording to be exact because it offers the benefit of not having to worry about gain staging while recording. It is not to record better sound quality (that's limited by the mics). It is to make the job easier and "fool proof" to yourself (and potentially harder for other people, but who cares, right?)



Yeah, but the 32 bit file is used to remove the need for gain staging while recording.

Yes, weeding through articles, it seems “32-bit recording workflow” is a more encompassing term. That it refers to new recording devices recording 32bit float (even if the situation and a ADC doesn’t get close to the theoretical limits of 32bit). I can see an easy application with videographers on location, who may not know how to set levels. But editing of those audio files is needed before delivery.

 

[gregorio]

Yeah, but the 32 bit file is used to remove the need for gain staging while recording.

Not really. We need to be a bit careful here because just setting the gain to zero and recording in 32bit wouldn’t work a lot of the time. With these new devices there is still gain staging, a great deal of it in fact (from what I can work out) but it’s done internally, with an arrangement of two differently calibrated ADCs per channel (and using digital attenuation on one), rather than the recordist having to do any gain staging. So although there are a lot of devices out there which will record 32bit files, going back many years, only the recently released ones (using this specific arrangement of ADCs and processing) can be used without gain.

TBH, the same thing could probably be achieved with 24bit but as the processing required to mix/splice/merge the two ADC outputs is done with a 32bit processor and as 32bit allows recovery from digital clipping, it’s easier just to write the 32bit output rather than convert it back to 24bit.

Here is you not understanding a camera records at 16bit, and due to exposure settings, doesn’t fill up that entire range!

I know, I’ve already stated that several times. I don’t understand what cameras are doing under the hood, I can’t make any informed comments, it’s way off-topic, it’s not sound science and, I’m not really interested anyway!

…you expect me to remember verbatim a previous post that goes half a page.

If it’s questions you yourself asked/posted and you’re going to insult me because I tried to answer them, then “Yes”, I do expect at least that!

G

 

[Davesrose]

I know, I’ve already stated that several times. I don’t understand what cameras are doing under the hood, I can’t make any informed comments, it’s way off-topic, it’s not sound science and, I’m not really interested anyway!

If it’s questions you yourself asked/posted and you’re going to insult me because I tried to answer them, then “Yes”, I do expect at least that!

G

It was a really simple post, with one paragraph that went over the similarity of a 16bit ADC that never truely reaches the full limits of 16bit in the recording stage. And before that, what started this whole exchange, was utterance of "32bit workflow" in relation to recording devices. Instead of just interjecting your experience, it's weeding through insults, taking the other poster's content out of context, first refusing to acknowledge there are 32bit recording devices, before getting to any content of how they work or what application they might be helpful in the recording stage. Now if I had first said "32bit recording workflow", I don't know if that would have triggered you....but somehow I think it would.

When it comes to these devices, it seems they attempt to make the noise floor redundant. Since they, in theory, have a file that can capture the S/N in any situation. One potential advantage is a sudden change in recording environment: https://www.thebroadcastbridge.com/content/entry/14755/hands-on-with-32-bit-float-recording . How that's processed by the device would be an interesting technical topic.

 

[gregorio]

first refusing to acknowledge there are 32bit recording devices

So you’re claiming that I’m refusing to acknowledge the existence of something which I first used about 15 years ago and which I’ve personally owned since 2011? That (and the rest of the paragraph) is obviously nonsense!

When it comes to these devices, it seems they attempt to make the noise floor redundant.

This assertion is false. To be true, the Zoom F6 would have to operate at zero degrees Kelvin and/or break the laws of physics. Looking at the specs, the maximum potential input dynamic range appears to be 131dB, which is slightly less than 22bits and 14dB less than 24bit. All the F6 could potentially do is make the digital noise floor redundant but it’s already way beyond redundant with 24bit. What’s not redundant is the acoustic noise floor and in some unusual conditions the noise floor of the mic but 32bit recording doesn’t have any effect on that.

Since they, in theory, have a file that can capture the S/N in any situation.

We already have that with 16bit, let alone 24bit.

https://www.thebroadcastbridge.com/content/entry/14755/hands-on-with-32-bit-float-recording

Oh dear, that’s even worse than the first one!

“Sudden loud sounds can cause clipping and quiet sounds can get lost in the noise.” - Sure, sudden loud sounds can cause clipping but not if you know how to use gain staging. And, quiet sounds can get lost in the noise but 32bit recording doesn’t affect that in the slightest.

“For me, and a lot of people who record audio in the field, this is a magnificent advance in audio technology.” - Yes, for writers, journalists and all sorts of people who don’t know much about audio and gain staging but record it regardless, it’s a magnificent, game changing advance, because now you can’t screw up the recording due to gain staging ignorance. You’ll now have to rely only on using the wrong mic or wrong mic placement to screw up the recording!

“However, the dynamic range captured goes from 144 dB up to essentially infinite - over 1500 dB!” - Captured from what, a microphone that defies the laws of physics, placed 10ft away from about 100 supernovas all happening simultaneously?

“32-bit floating point processing certainly has great benefits in post.” - What great benefits does it have over the 64bit float processing we’ve been using in post for the last decade?

“Also, think of music recording. The 32-bit floating record mode allows mixers to do hands-free mixing of small music groups without worry of distorting the sound. That, alone, is revolutionary.” - For someone who’s never mixed a music recording before it might be revolutionary but for those who do it for a living, it would be a joke! At least until they invent AI that can replace mix engineers.

“Since the beginning of my career, I’ve also been paranoid about getting the correct levels … As one who has been using audio mixers for more than 50 years, there is no going back for me.” - Well that explains it and to be fair, it’s not as uncommon as it should be. 50 years ago and even 30 years ago when analogue recording was still common/standard, you absolutely had to be paranoid about getting the correct levels and the correct level was “as high/hot as possible without clipping”. The same rule of thumb was often true when recording in 16bit, though it had quite a bit more headroom. So, if you’re quite close to the clipping/distortion level, anything unexpectedly loud will clip/distort. That rule of thumb does not apply to 24bit though, 10dB of headroom or 30dB, no difference, you’re still miles away from the noise floor. Nevertheless, I still quite often see recordists leaving only 6-10dB headroom when recording in 24bit.

G

 

[Davesrose]

So you’re claiming that I’m refusing to acknowledge the existence of something which I first used about 15 years ago and which I’ve personally owned since 2011? That (and the rest of the paragraph) is obviously nonsense!

This assertion is false. To be true, the Zoom F6 would have to operate at zero degrees Kelvin and/or break the laws of physics. Looking at the specs, the maximum potential input dynamic range appears to be 131dB, which is slightly less than 22bits and 14dB less than 24bit. All the F6 could potentially do is make the digital noise floor redundant but it’s already way beyond redundant with 24bit. What’s not redundant is the acoustic noise floor and in some unusual conditions the noise floor of the mic but 32bit recording doesn’t have any effect on that.

We already have that with 16bit, let alone 24bit.

Oh dear, that’s even worse than the first one!

“Sudden loud sounds can cause clipping and quiet sounds can get lost in the noise.” - Sure, sudden loud sounds can cause clipping but not if you know how to use gain staging. And, quiet sounds can get lost in the noise but 32bit recording doesn’t affect that in the slightest.

“For me, and a lot of people who record audio in the field, this is a magnificent advance in audio technology.” - Yes, for writers, journalists and all sorts of people who don’t know much about audio and gain staging but record it regardless, it’s a magnificent, game changing advance, because now you can’t screw up the recording due to gain staging ignorance. You’ll now have to rely only on using the wrong mic or wrong mic placement to screw up the recording!

“However, the dynamic range captured goes from 144 dB up to essentially infinite - over 1500 dB!” - Captured from what, a microphone that defies the laws of physics, placed 10ft away from about 100 supernovas all happening simultaneously?

“32-bit floating point processing certainly has great benefits in post.” - What great benefits does it have over the 64bit float processing we’ve been using in post for the last decade?

“Also, think of music recording. The 32-bit floating record mode allows mixers to do hands-free mixing of small music groups without worry of distorting the sound. That, alone, is revolutionary.” - For someone who’s never mixed a music recording before it might be revolutionary but for those who do it for a living, it would be a joke! At least until they invent AI that can replace mix engineers.

“Since the beginning of my career, I’ve also been paranoid about getting the correct levels … As one who has been using audio mixers for more than 50 years, there is no going back for me.” - Well that explains it and to be fair, it’s not as uncommon as it should be. 50 years ago and even 30 years ago when analogue recording was still common/standard, you absolutely had to be paranoid about getting the correct levels and the correct level was “as high/hot as possible without clipping”. The same rule of thumb was often true when recording in 16bit, though it had quite a bit more headroom. So, if you’re quite close to the clipping/distortion level, anything unexpectedly loud will clip/distort. That rule of thumb does not apply to 24bit though, 10dB of headroom or 30dB, no difference, you’re still miles away from the noise floor. Nevertheless, I still quite often see recordists leaving only 6-10dB headroom when recording in 24bit.

Here's another example of taking my posts out of context and going into your own stream of consciousness to now get into insults with an author who says he has 50 years of field recording. You very clearly were dismissing 32bit recording devices that are new compared to what you were using 15 years ago (which although might have recorded in 32bit, were not processing to not need gain control as a function of time during the full recording session). Refusing to first read through and acknowledge other fields, such as the above article from a broadcast resource.

 

[gregorio]

Here's another example of taking my posts out of context and going into your own stream of consciousness

And again, you post an article as supporting evidence, I explain why it’s nonsense and doesn’t support your argument and you throw your toys out of your pram and make ad hominem attacks?!

… an author who says he has 50 years of field recording.

Yes, there are many journalists, one man ENG reporters and others who use field recorders routinely for decades but know very little about audio, which is not really surprising as that’s not their job. And, they can be particularly ignorant if they were first taught/shown how to use a field recorder before 24bit recording was standard.

You very clearly were dismissing 32bit recording devices that are new compared to what you were using 15 years ago

But I gave an example pages ago (post #3,452) that explained a potential benefit of 32bit audio file recording, which I’ve used on occasion over the last decade. What I dismissed is the notion that 32bit recorders can actually capture 32 bits worth of audio data. I’m skeptical that they are a better or even as good a solution as standard 24bit recorders in all those TV/Film situations where field recorders are employed (by professionals) but probably would be better for most amateurs.

G

 

[Davesrose]

Your post was a very clear stream of consciousness attacking the article's author about his experience level, before at the end realizing he has 50 years experience (and rationalizing why "it's a game changer" for him might be bunk). You go ad hominem about how people who are using these new 32bit recorders are ignorant who don't know any intricacies of recording or the implications of mixing. Now sure, they are more practical with productions that have a videographer who isn't monitoring the audio. But there are also audio professionals using them. You can't accept that we understand a 16bit ADC does not really fill the full range of a 16bit file, or that 20bit ADCs don't approach anything near 32bit with the file they output from processing. I find it impossible to have any productive exchange with you. It would be refreshing if we can accept that these devices exist, why they are used, how they work, and if there's an implication of those automatic adjustment of those files in the editing. But no, we have to go in circles with semantics and lengthy replies that didn't first read the post, and begin with ad hominems.

 

[bigshot]

What causes people to grab on like this? I'm genuinely curious. There's a whole level of miscommunication going on here that has absolutely nothing to do with sound science, yet we see it here all the time. What is it that causes people to act like this? They'd never do this in the real world. No one would want to be around them. Is it an internet thing? I don't see this sort of thing as much in other Internet forums. Something is attracting it here. Maybe it's the ban on discussion of ABX in the rest of Head-Fi. People think this is the argument forum, not the science forum.

 

[Davesrose]

G

What causes people to grab on like this? I'm genuinely curious. There's a whole level of miscommunication going on here that has absolutely nothing to do with sound science, yet we see it here all the time. What is it that causes people to act like this? They'd never do this in the real world. No one would want to be around them. Is it an internet thing? I don't see this sort of thing as much in other Internet forums. Something is attracting it here. Maybe it's the ban on discussion of ABX in the rest of Head-Fi. People think this is the argument forum, not the science forum.

Good question, as I'm one of the guilty parties. I think part of it is a miscommunication on the onset, and need to respond to a perceived attack. Maybe a large part of it is not understanding full subtle context on an internet forum, or not paying attention to the person posting the way you'd listen to them in a real social setting. There could have been just nice informative info about these new 32bit recording devices, and how they effect recording and editing in a production with "32-bit recording workflow". But since we're mired by miscommunication, I'm letting go. Sorry to other members.