24bit vs 16bit, the myth exploded!

[bigshot]

What kind of drum roll are we referring to? The kind in the circus right before someone jumps off a high ladder into a net?
I personally do not find that typical drum roll to be either very loud or all that dynamic.

That's because it's halfway across a circus tent from you. Put it at the end of your arms and it's an entirely different story.

1. Almost all music recorded in 24 bits.

I'm just going to address this one because it's all I need to address...

Music is recorded with 24 bits of dynamic range so there's plenty of latitude to adjust levels in the mix. But the final mix doesn't include even a fraction of that dynamic range because all sorts of compression is applied to make the music comfortable to listen to. This is a good thing. Too much dynamic range is irritating and forces you to keep reaching for the volume control to turn up quiet parts to make them audible, and turn down loud ones so they don't blast your ears. This compression can be electronic in the form of compressors and limiters, or it can be acoustic in the form of miking from a distance and allowing hall ambiences to soften the dynamic peaks. This process is the same for pop music, jazz and classical... just to different degrees. All music is balanced to sit comfortably within the ear's natural dynamic range of about 45 to 50dB at a time. That's why LPs can have a dynamic range of less than 50dB and still sound good. That's why we never hear the noise floor of a CD without jacking the volume control on the fade outs. Music is mixed to sound good to ears. It isn't mixed to conform to abstract numbers relating to the extremes of human perception. If you want to recreate the sound of a jackhammer turning on and off in the depths of Carlsbad Caverns, then you are going to need 24 bits to do that. But if you want to recreate the sound of a symphony orchestra or rock band, redbook is perfectly capable of doing that. In fact, redbook has more dynamic range than you need, and that is proved by some overly dynamic recordings on the BIS label that are a chore to listen to.

CD sound is all you need. See the link in my sig.

That said, I have a lot more respect for people who make and record music than I do people who just enjoy posting in forums about recording quality. It's all about the music. That is all. Do not pass Go. Do not collect $200. Music is the purpose of all this. The equipment is just the means to the end. Great music with great sound is great. But great music doesn't get any greater by including frequencies only bats can hear and noise floors that reveal the heartbeat of the guitar player between songs.

I have actually read a lot of AES papers, and they're almost universally really bad research

If you want to participate in peer review you have to work to become a peer. It isn't enough to be an armchair quarterback. You have to get out there and actually produce something. I think if you were more involved in the career of music recording, you might have more understanding of the way music should be reproduced and perceived. Those aren't totally separate things. They're inter-related on a million different levels. It would help if you listened to other people and made an effort to understand what's being said to you. It would also help if you stuck to honest discourse without letting your ego make you wander into logical fallacies and argumentative tactics that obscure the truth instead of revealing it. Just a suggestion. Feel free to ignore it if you want.

 

[reginalb]

Looking at Amirm's graph here, I think I finally have a good grasp of what he's arguing and that it's impractical at best.

The key shows what on close inspection appear to be an open circle for classical, an open box for jazz, an open triangle for rock, and upside down triangle for others.

Then they're filled in if there is electronic augmentation.

In the section above 120dB there are I believe 6 rock concerts, 5 "Others", 3 or 4 jazz and 0 or 1 classical.

I include this summary because that is one hard chart to read and welcome corrections to what I'm seeing here.

I guess the argument is that we are pretending that noise shaping is not a thing, and thus 96dB is what we have to work with, and 129-96 is only 33dB, which isn't low enough and therefore the full dynamic range of 24-bits is necessary. So if I listen to my music so loud that my ears ring when I leave (which they do at the average rock concert) and somehow I'd hear the noise floor over the ringing in my ears, then I would need 24-bits (assuming that noise shaping doesn't exist, which it totally does) for proper reproduction of sound. All this because it has to be "perfect," and by "perfect" we mean, "Damaging to our ears." And again, I don't think I'd hear the noise floor at all, because my ears would be ringing.

And yes, Amir, it would damage our ears. Exposure to sound at the level that causes temporary hearing loss will, over time, cause long term damage to our ears. Since you're obsessive about citations, here you go:
http://www.jneurosci.org/content/29/45/14077.short

Here's some more reading on the affects of levels as low as 96dB average (in clubs) on DJ's, listening to music mind you, not "impact noise" which you seem to feel is much more harmful than musical noise:
https://www.cambridge.org/core/jour...n-nightclubs/573BAE71AFF87AB06570E5991E5AC6CA

 

[reginalb]

"My ears ring at the end of a concert
if I don't wear hearing protection."


Then something is not being done correctly at that concert if it's that loud. It's not a Nascar race.

I agree completely, but it is a near universal experience for me unless I wear hearing protection (which I do). My wife likes to be near the speakers (she is actually hard of hearing) so that has something to do with it, I'm sure. But if you need more proof, just look at those peak levels from Amir's citation. They're insane.

 

[RRod]

Looking at Amirm's graph here, I think I finally have a good grasp of what he's arguing (and that it's insane).

The key shows what on close inspection appear to be an open circle for classical, an open box for jazz, an open triangle for rock, and upside down triangle for others.

Then they're filled in if there is electronic augmentation.

In the section above 120dB there are I believe 6 rock concerts, 5 "Others", 3 or 4 jazz and 0 or 1 classical.

I include this summary because that is one hard chart to read and welcome corrections to what I'm seeing here.

I guess the argument is that we are pretending that noise shaping is not a thing, and thus 96dB is what we have to work with, and 129-96 is only 33dB, which isn't low enough and therefore the full dynamic range of 24-bits is necessary. So if I listen to my music so loud that my ears ring when I leave (which they do at the average rock concert) and somehow I'd hear the noise floor over the ringing in my ears, then I would need 24-bits (assuming that noise shaping doesn't exist, which it totally does) for proper reproduction of sound. All this because it has to be "perfect," and by "perfect" we mean, "Damaging to our ears." And again, I don't think I'd hear the noise floor at all, because my ears would be ringing.

And yes, Amir, it would damage our ears. Exposure to sound at the level that causes temporary hearing loss will, over time, cause long term damage to our ears. Since you're obsessive about citations, here you go:
http://www.jneurosci.org/content/29/45/14077.short

Here's some more reading on the affects of levels as low as 96dB average (in clubs) on DJ's, listening to music mind you, not "impact noise" which you seem to feel is much more harmful than musical noise:
https://www.cambridge.org/core/jour...n-nightclubs/573BAE71AFF87AB06570E5991E5AC6CA

Are there rock concerts where the ambient noise levels are much around 33dBSPL? Also, what are the crest factors here. I can believe a 130dB peak if the crest is 40dB, but do I really expect that at a rock concert? The numbers just add up to way too high an average listening level to be at 'optimal' positions, and even then, there is crowd noise...

 

[bigshot]

"My ears ring at the end of a concert
if I don't wear hearing protection." Then something is not being done correctly at that concert if it's that loud. It's not a Nascar race.

The level of attention to response curves in amplified concert venues is often dismal. They sometimes depend too much on automated calibration that leaves big narrow spikes of level imbalances in the center of the most sensitive hearing range. If the spike is high enough and narrow enough, it isn't very audible. But it can wreak havoc on your eardrums. I've experienced this too many times at large rock shows, especially at outdoor arenas where they don't figure they need to tightly control the response.

 

[amirm]

For the actual content, THANK YOU. You finally took something and actually responded accordingly. This is the first post I've read where you actually directly addressed something, and it was actually enlightening. I learned what I already know, peak levels at rock concerts are WAY too loud. I wear hearing protection at rock concerts. I apologize if I missed the information about the mic, but when this direct criticism was levied you absolutely DID NOT present this in response. You literally responded by saying that the paper has 63 citations at the end (meaningless), and then the author's bio (also meaningless).

Rock concerts are but one category represented in that research. And I provided the same answer a week ago: https://www.head-fi.org/threads/24bit-vs-16bit-the-myth-exploded.415361/page-306#post-13864815

This is from the same post where I talked about the 67 references which sounds like you had read and hence your complaint about it:

We have a Jazz unamplified sample nearing 130 db SPL.

I wish they would have chosen a different shape (or at least two that are more discernible in reproduction if they were different) for jazz and classical.

I have quoted text which makes much of this clear. Again from the above post:

And from the paper I quoted earlier from *other research* than Fielder's

As you see even solo piano can have peak level of 103 db SPL. So please don't twist the data into it being about rock concert. It is improper to spin data this way.

 

[amirm]

With regards to bad research, I'm not going to go down the list, but there are a ton of pseudo science papers published in this field, and quite a few in that particular paper. Typically low sample sizes, poor documentation of methodology, and questionable conclusions have been the hallmark of papers I've read in this community.

You read in the "community?" In other words you personally have not read of the AES Journal papers? Just going by what is said on forums? And based on that damned everything published by J.AES, J. ASA, etc.?

 

[reginalb]

You read in the "community?" In other words you personally have not read of the AES Journal papers? Just going by what is said on forums? And based on that damned everything published by J.AES, J. ASA, etc.?

Nope, I've read papers that don't cost money to access. Which would be silly for me to do given that I don't work in audio. It would be a waste of money.

Rock concerts are but one category represented in that research. And I provided the same answer a week ago: https://www.head-fi.org/threads/24bit-vs-16bit-the-myth-exploded.415361/page-306#post-13864815

This is from the same post where I talked about the 67 references which sounds like you had read and hence your complaint about it:

We have a Jazz unamplified sample nearing 130 db SPL.


I have quoted text which makes much of this clear. Again from the above post:

And from the paper I quoted earlier from *other research* than Fielder's




As you see even solo piano can have peak level of 103 db SPL. So please don't twist the data into it being about rock concert. It is improper to spin data this way.

103 dB would be easily contained within the dynamic range of CD, even without noise shaping....

I didn't twist the data, I literally counted out each genre and the number that were over 120 dB. How about you stop misrepresenting everything I say?

The level of attention to response curves in amplified concert venues is often dismal. They sometimes depend too much on automated calibration that leaves big narrow spikes of level imbalances in the center of the most sensitive hearing range. If the spike is high enough and narrow enough, it isn't very audible. But it can wreak havoc on your eardrums. I've experienced this too many times at large rock shows, especially at outdoor arenas where they don't figure they need to tightly control the response.

I have personally found small venues that are essentially bars to be the worst offenders. But have had the problem at outdoor venues as well.

 

[amirm]

Are there rock concerts where the ambient noise levels are much around 33dBSPL?

33 db SPL what? I have explained so many times that when it comes to minimum noise level, you must examine the spectrum. https://audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

But no, I don't know what the SPL level is at some random concert. The studies I have shown you represent the noise level in the same halls where peak measurements were performed such as Davies hall above.

 

[bigshot]

As you see even solo piano can have peak level of 103 db SPL.

If you shove the microphone right up next to the hammers with the lid down it would definitely register 103dB in full forte. After all the piano is a percussion instrument. It's capable of loud transients. But that isn't generally how you mike a piano. You normally want some of the woody body of the sounding board. Usually I've seen pianos miked with the lid closed a little above the performer's head at 45 degrees down, or with the lid up at a distance of a few feet from the harp to the side. This takes the curse off the loud percussive "thump" and gives a more musical sound. Miked this way, I doubt if the peaks would get much above 85dB. Ideally the performer would modulate his dynamics with pedaling and touch to keep within a comfortable level and wouldn't necessarily play balls out (Jerry Lee Lewis excepted of course!) This is another case where the instrument is capable of producing a very loud volume in theory, but when it comes to practice, music just doesn't call for that sort of thing because it's uncomfortable to listen to.

In a live performance in a concert hall, the room ambience would suck up a lot of the dynamics. The audience would be getting something in the range of normal music... a dynamic range of 50dB or so. The idea of recorded music is to simulate that sort of experience, so the various microphones would be mixed together at levels that give a pleasing dynamic, even though the natural dynamics close miked would be off the chart.

I have personally found small venues that are essentially bars to be the worst offenders. But have had the problem at outdoor venues as well.

oh yeah! tight enclosed space... sloppy work at the board... a band that is used to playing larger venues... that's a recipe for painful ears.

 

[RRod]

33 db SPL what? I have explained so many times that when it comes to minimum noise level, you must examine the spectrum. https://audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

But no, I don't know what the SPL level is at some random concert. The studies I have shown you represent the noise level in the same halls where peak measurements were performed such as Davies hall above.

The thrust of my post was questioning the levels of the rock data given an optimal listening position. I understand that the 130dB peak for orchestral material would probably be under more stringent noise conditions.

 

[sonitus mirus]

We have a Jazz unamplified sample nearing 130 db SPL.

$10 is a lot to me. I don't have money to burn to spend on an AES subscription. I appreciate your willingness to present some of the data from papers that I would not otherwise have an opportunity to read. How often did these peaks occur and for what duration? Is there any evidence that these occurrences were audible? I don't need an out of context example, I mean specifically in the situations where these peak levels were recorded?

Nice to see that a majority of the classical genre has peaks that appear more reasonable.There were no occurrences recorded over 120 dB with any of the classical concerts that were measured. 90 dB peak? Was it some elementary school student's solo?

 

[71 dB]

Almost 5000 posts in this thread. In my opinion the first three pages of discussion board threads are the most fruitful and after that they seem to always become endless fights were people are hitting each other from their foxholes. For me 16 bit dynamic range has always been enough and will be in the future. Those who for some reason suffer from the noise floor of 16 bit audio have 24 bit downloads. Maybe they are even offered 64 bit files in the future and have nearly 400 dB of dynamic range to listen to exploding stars at realistic levels in the silence of Skywalker Scoring Stage. The music I want to ever listen to (at my noisy home, not at Skywalker Ranch) fits in 16 bits just fine. Heck, I'd manage with 14 bits only!

 

[gregorio]

I've started to bring my earplugs even to the theater nowadays just in case it's one of those movies.

The maximum peak in the middle of a cinema is going to be about 115dB SPL. However, that assumes the standard calibration level (called "Dolby 7.0") and a vast number of cinema screens actually use a lower level setting due to audience complaints. Dolby 5.5 is very common and results in peak levels in many cinemas being lower than 110dB SPL.

Gregorio mentioned 60dB as something we would almost never pass on recorded albums. I have a few tracks going beyond so it's not a strict limit for sure. but it was never intended to be ...

Absolutely! There are literally millions of commercial recordings, the vast majority have a dynamic range of around 50dB or less. A tiny fraction have a dynamic range of 60dB and of those, a very tiny fraction have a dynamic range greater than 60dB. I don't believe any one knows for sure but possibly only a handful of those millions.

1. Almost all music recorded in 24 bits.
2. Conversion to 16 bits (without dither and noise shaping) chops off 8 bits of that. Let's be generous and say that half of that was noise anyway. So that leaves us with 4 bits thrown away. That represents 4*6 = 24 db increase in noise floor.
[4] But let's say we did use noise shaping. What that translates into is that we took perfectly good 24 bit music file, added 24 bits of noise to it, and then applied a transformation to that noise. Why do we want to do that?
[4a] Why not just give the original 24 bit file to people who want it?
[5] This is on top of the fact that it is not up to you and I to apply noise shaping. It is up to the entire content production industry, vast majority of which have no background in signal processing and you will be lucky if you get dither for conversion of 24 to 16 let alone noise shaping.
[6] So in summary, conversion to 16 bits from 24 bits is a form of lossy compression. Noise shaping makes it a perceptual technique so no different than lossy codecs. We no longer have a need for such a conversion.
[7] Give the bloody 24 bit files and let's be done with it! ... Really folks, physical formats that forced these things on us are gone. We have total freedom. Let's offer the consumer the option to download music at originally captured rates.
Technology and big companies are no longer a barrier to us doing so. Let's celebrate our newly found freedom! I know I am....

1. NO, no music is recorded at 24bits, it's recorded in a 24bit container but no ADC achieves more than about 20bits performance.
2. No, we've only got about 20bit performance to start with, then we need headroom! And on top of that we've got the recording noise floor; environmental noise (room, musicians, etc.) and equipment noise, SNR of the mics + mic-preamps. So in practice, "almost all music is recorded"at no more than about 18bits and typically fewer!
4. No, that is NOT true! We do not apply noise shaping to the original 24bit container tracks, we ONLY apply noise shaping as the final step in mastering and the music at that stage is typically at 64bit float.
4a. I asked you before and you ignored the question! You want the original recorded channels or do you want the music actually mixed and mastered? Not that it really matters because we are NEVER going to give you the originally recorded channels!!
5. No it's not. Noise shaped dither is applied as the final mastering step, therefore the only person who applies it is the mastering engineer and NOT the "entire content production industry"!
6. I disagree that applying noise shaped dither is the same as a lossy codec but even taking your weird definition of "lossy", there is NO option as even 24bit is effectively "lossy" because the music is created in 64bit float! Whatever container we distribute the music in it will always be "lossy", according to your definition!
7. Again, you cannot have the "originally captured" rates/channels, commercial artists will only sell you a finished product. So, continuing to scream and shout for something you cannot have comes across as nothing more than the ravings of one of those extreme audiophile nutters!

You can have a 24bit file which has been truncated twice or you can have a noise-shaped 16bit file. The reason for this thread is; what are you gaining with the double truncated 24bit file which makes it audibly superior to the noise-shaped 16bit file and worth the higher price of high resolution (which is not in effect any higher resolution)? The problem, as has been explained to you, is that going down the road of higher price for something only marketed as hi-res is the FREEDOM to get ripped-off by companies offering ever higher resolution for content and equipment which is NOT higher resolution. That's why we're starting to see 768/32!

Surely we agree that if live performances resembled occupations noise in constructions or factory floor, people would not go pay and sit there to listen to them for one hour or more.

Of course we cannot agree that! Even governments don't agree with that, as many have now passed laws restricting the SPL levels of music performances on the grounds of health & safety and, I doubt there are many here who have not at some time in their life paid to go to a club/gig and not suffered hearing symptoms (tinnitus or threshold shift for example). Why are you ignoring the supplied scientific data of hearing damage caused by MUSIC?

33 db SPL what? I have explained so many times that when it comes to minimum noise level, you must examine the spectrum.

Then WHY DON'T YOU? Why are you steadfastly refusing to do EXACTLY what you are accusing others of not doing? Why???? YOU quoted the highest peak of a live gig was 129dBSPL, so, what was the noise floor and it's spectrum at that live gig? Are you saying that the noise floor of that gig in the critical band was less than 9dB SPL, which it would HAVE TO BE in order for the 120dB of dynamic range with 16bit to be insufficient. If you are saying that, where's your evidence?

All the above (and more) has been explained to you before, you just ignore it and just keep repeating the same INCOMPLETE and/or INAPPLICABLE data/studies. For this reason, you have now firmly entered the realm of trolling!!

G

 

[reginalb]

...YOU quoted the highest peak of a live gig was 129dBSPL, so, what was the noise floor and it's spectrum at that live gig? Are you saying that the noise floor of that gig in the critical band was less than 9dB SPL, which it would HAVE TO BE in order for the 120dB of dynamic range with 16bit to be insufficient. If you are saying that, where's your evidence?

...G

You left out an extremely important part of Amir's conditions, which is that noise shaping isn't a thing, so you can only say 96dB of dynamic range, not 120. So we're looking at a 33dB noise floor.