PHILIPS RED BOOK

[oivavoi]

But isn't Amir's point rather straightforward, though? I'm not sure we need neither simulations nor blind tests here. It's a simple question of logic. In acoustic music settings, min to max dynamic range may reach 110-120 db in the presence region (there will always be a higher noise floor in the bass region). I can attest to that, from my experience of singing in a semi-professional choir which frequently performs with orchestras: Within the same piece, it can go from very very low - whispering at a level which is barely audible - to super loud. I've occasinally snuck out my iPhone with a good quality SPL meter app during practice, and I've measured peaks of 118-119 db from where I stand among the basso. This kind of dynamic range rarely gets reproduced in the home, neither by our systems nor by the recordings.

So assuming one has:
a) a very quiet room
b) a system which is capable of playing extremely loud
and
c) electronics with a very low noise floor,

I can see the point of hi-rez recordings with a dynamic range of up to 120 db. 120 db is not continuous, of course, it's about very short peaks.

If either a, b or c doesn't apply, I can't see any point in hi-rez. And even 90 db of dynamic range can be plenty loud, I think. But for the ultimate of fidelity? Sure, why not.

 

[Wombat]

The ears and mind introduce individual differences in subjects between what is heard along with personal subjectivity. Hardly identical sound processing giving the same auditory result for all listeners in a test.

 

[Cosmik]

But isn't Amir's point rather straightforward, though? I'm not sure we need neither simulations nor blind tests here. It's a simple question of logic. In acoustic music settings, min to max dynamic range may reach 110-120 db in the presence region (there will always be a higher noise floor in the bass region). I can attest to that, from my experience of singing in a semi-professional choir which frequently performs with orchestras: Within the same piece, it can go from very very low - whispering at a level which is barely audible - to super loud. I've occasinally snuck out my iPhone with a good quality SPL meter app during practice, and I've measured peaks of 118-119 db from where I stand among the basso. This kind of dynamic range rarely gets reproduced in the home, neither by our systems nor by the recordings.

So assuming one has:
a) a very quiet room
b) a system which is capable of playing extremely loud
and
c) electronics with a very low noise floor,

I can see the point of hi-rez recordings with a dynamic range of up to 120 db. 120 db is not continuous, of course, it's about very short peaks.

If either a, b or c doesn't apply, I can't see any point in hi-rez. And even 90 db of dynamic range can be plenty loud, I think. But for the ultimate of fidelity? Sure, why not.

I have no objection to extra bit depth if it doesn't cost very much to do it. But supposing it could be shown that polar bears and pacific islands might perish because of the energy needed to handle the extra data..?

In the report, I think it says that 16 bit with noise shaped dither can approach 115dB, anyway..? Also this:

Thus, 16 bit audio can go considerably deeper than 96dB. With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice

 

[oivavoi]

I have no objection to extra bit depth if it doesn't cost very much to do it. But supposing it could be shown that polar bears and pacific islands might perish because of the energy needed to handle the extra data..?

In the report, I think it says that 16 bit with noise shaped dither can approach 115dB, anyway..? Also this:

Hehe, I don't want to be known in the history books as the audiophile who made polar bears go extinct!

If CD-quality indeed can approach a dynamic range of 110-115 db, the argument for hi-rez loses much of its validity, as I see it. But these technical issues are extremely far from the level of my actual competence and insight when it comes to digital technology.

 

[bennetng]

Actually I am unwilling to do such listening test in the original SPL, even when getting hired or paid. I found some concerts and arcade game centers too loud and don't want to stay too long in those places.

A little girl screams in front of me is already painful enough.

 

[oivavoi]

Actually I am unwilling to do such listening test in the original SPL, even when getting hired or paid. I found some concerts and arcade game centers too loud and don't want to stay too long in those places.

A little girl screams in front of me is already painful enough.

My experience is that loud acoustic music in good venues often is less fatiguing to the ears than reproduced music in the home, for some reason. That said, I also try to take care of my hearing, and my baseline listening level in the home is usually between 74 and 80 db. Not very loud. But much of the music I listen to has large dynamic peaks, which I've measured to be as loud as 110 db. Doesn't seem like my present monitors are able to go any louder. I've never perceived these peaks to be too loud, probably because they're so short that they almost go unnoticed.

 

[RayDunzl]

Indeed. So if I go from a maximum overlap of three percussive impulses, say, up to 100, that's approximately six doublings, equalling an extra 36dB above what I thought my maximum was - in some freak circumstance that happens every million years.

Maybe...

But each doubling of coherent sources is only 3dB over the incoherent doubled sources case (for 18dB instead of 36dB), assuming your 100 percussions are being played aligned (freak case) versus 3 aligned and 97 unaligned simultaneous percussions (more normal case).

 

[Blumlein 88]

Hehe, I don't want to be known in the history books as the audiophile who made polar bears go extinct!

If CD-quality indeed can approach a dynamic range of 110-115 db, the argument for hi-rez loses much of its validity, as I see it. But these technical issues are extremely far from the level of my actual competence and insight when it comes to digital technology.

Okay, these are digital only files done with Audacity. I have done such things with real gear, and the dither works up to the limits of the electronics noise of the gear.

First I created a 2 khz tone at -118 dbFS. You see here in this 1024 bin FFT what it looks like in 32 bit float.

Next we have this file saved as 16 bit without dither applied. You can see the signal disappeared as it fell below the LSB.

So now I apply dither when saving it as a 16 bit file. In this case TPDF dither. And you'll see there it is. The level shown has raised a fraction of a db as it is only a few db above the noise floor of the dithered LSB.

Here is the same file this time saved as 16 bit with shaped dither. An even lower noise floor below 5 khz in which a signal could be recovered.

Finally the same file saved as dithered 24 bit.

So you can record and recover a noisy signal at -120 db with dither. It isn't the same as the signal at 24 bit which is less noisy (though still some noise from the thermal noise limits of the gear in use). We can hear some 10-15 db into noise floors at least. So I am not sure you could hear this on a playback system set to play 0 db FS at 120 dp SPL. It might be hard to pick up in the noise using TDPF dither. You can hear such a thing somewhere around -110 dbFS perhaps. If shaped dither were used, you might hear it.

So a simple un-monkeyed with dithered 16 bit has a human perceivable 110 db or so of dynamic range. One with shaped dither might manage over most of the range more like -120 db. Finding such recordings will be difficult. It does show that if one did all work at 24 bit, and did one final downsampling with dither to 16 bit you likely are getting dynamic range equal to or exceeding nearly all playback systems.

 

[bennetng]

http://audio.rightmark.org/lukin/dither/dither.htm
The author Alexey Lukin, was one of the developers of Rightmark audio analyzer (RMAA) and currently working in iZotope.

 

[Blumlein 88]

I took the dithered 16 bit files, amplified them 60 db so you could hear them. That makes this previously -118 db 2 khz tone now -58 dbFS. I am attaching them as VBR mp3 files.

The TPDF dither doesn't make hearing the 2 khz tone easy. The shaped dither even though total dither level is 10 or so db higher makes it much easier to hear the 2 khz tone.

It is a zip file that will open into a pair of mono mp3 files of 10 seconds each. So you can hear it for yourself.

EDIT:I added another file. This one is -110 db 2 khz tone with TPDF dither with 60 db of gain. You can begin to hear the tone though still not as clearly as in the shaped dither version.

 

[watchnerd]

I think the paper kind of demonstrates the point that CD is pretty much adequate, in that it has to get every link in a long chain at the absolute limits of what is possible in order to get the 120dB requirement. The reality will be that 99.999% of recordings coupled with listeners and their listening environments won't come anywhere near it.

Not even that many, given the use of compression in the mixing/mastering. Orchestral works are usually less compressed than other genres, but it's still used to allow home listeners to set a comfortable level between 'soft' and 'loud' without having to run for the volume knob.

 

[Blumlein 88]

That is the study from Fielder which shows that to not be a limitation. I wrote a digest of his research for the playback side in this article: https://audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

In a nutshell, we need to look at the spectrum of noise. When we do, the SPL noise numbers that we think are too high, turn out to not be the case where our hearing is most sensitive.

Furthermore, we are able to hear noise coming out of speakers or a band playing, well below room noise. The directionality of the noise that way is key, versus diffused one in the environment.

He also examines the limits of microphones, speakers, amplifiers, etc. The paper is quite long, is peer reviewed, and has some 60 references. It is very hard to shoot holes in it.

Here are some sample data from his paper, Dynamic-Range Issues in the Modern Digital Audio Environment*
LOUIS D. FIELDER, AES Fellow

View attachment 9735

So here we see despite these recording venues having noise floors at low frequencies as high as 45 db, they are all pretty close or beat threshold of hearing. They do so because their noise level is as low at 0 db or even lower at mid-frequencies where we are most sensitive.

This is peak playback levels in same and other venues across larger number and genre of performances:

View attachment 9736

So we see that peak is above 120 db even for non-amplified music in some cases (black bars). So if we take 0 db for the lowest level and 120 db for highest, we arrive at 120 db dynamic range requirement.

Here we see that threshold of hearing white noise -- the kind created by equipment/channel -- is actually lower than threshold of hearing that is based on detection of tones:

View attachment 9738

Here is the data on microphones:

View attachment 9739

On and on. Here is the abstract summarizing it:

View attachment 9740

Why the ugly red background. YUCK!

Good paper. It does note that most of the sound causing up to 129 db SPL peaks is below 1 khz and mostly 500 hz. As we have raised thresholds in those areas, one could get by with less dynamic range for those frequencies. The J17 suggested pre/de emphasis curve is interesting as it would take advantage of this fact netting us better dynamic range where we can hear it.

One might conclude in the 3-7khz range signals of 120+ db aren't encountered and they usually aren't. Though I have recorded a few unusual percussion instruments that hit 110 db SPL or better in that range.

 

[watchnerd]

I think you touch on a lasting factor at the end of this, there’s a academic advantage to high res we can argue about till the cows come home but really unrestrained by physical media the question is why not high res?

Can it do harm? Is it that important?

There’s a responsibility for us to not stoke the fear in the audiophile by claiming some huge actual advantage over Red book imo I spin cd’s with a consideration for all the mitigating circumstances involved in play back Im certain I’m not missing out on anything. The marketing of high res is cynical imo maybe that sticks into a few of the red book forever types ( like me).

Can you show the current implementation (available recordings ) of high res brings a advantage? Not some lovely lab sample for a experiment , the tracks available over streaming services or downloads? From what Iv seen these tracks are often not what they appear to be and certainly the potential advantage seems lost.

Regardless, all of this came about because we had to comply with a fixed format at 16/44.1. When distributing music online, there is no reason whatsoever to try to conform it to CD specs. Music should be released it its originally captured bit depth and sample rate. People can make their own conversion to 16/44.1 if necessary and so can the music distributor.

Monty says high resolution is silly to wasteful. I'm a bit surprised to see some here jumping on the high res bandwagon when many probably can't ABX standard lossless vs high resolution:

https://people.xiph.org/~xiphmont/demo/neil-young.html

 

[bennetng]

I took the dithered 16 bit files, amplified them 60 db so you could hear them. That makes this previously -118 db 2 khz tone now -58 dbFS. I am attaching them as VBR mp3 files.

The TPDF dither doesn't make hearing the 2 khz tone easy. The shaped dither even though total dither level is 10 or so db higher makes it much easier to hear the 2 khz tone.

It is a zip file that will open into a pair of mono mp3 files of 10 seconds each. So you can hear it for yourself.

EDIT:I added another file. This one is -110 db 2 khz tone with TPDF dither with 60 db of gain. You can begin to hear the tone though still not as clearly as in the shaped dither version.

Yes. With non-music single tones it is even possible to achieve lower values. I attached a -125dBFS 2khz tone encoded to 16-bit with 60dB gain.

The noise shaper of this and my previous post is from foobar2000's file converter, as it is free everyone can try it for themselves. The noise shaping is more aggressive than yours and thus the higher waveform peaks with a +60dB gain.

 

[Blumlein 88]

Monty says high resolution is silly to wasteful. I'm a bit surprised to see some here jumping on the high res bandwagon when many probably can't ABX standard lossless vs high resolution:

https://people.xiph.org/~xiphmont/demo/neil-young.html

I would prefer to hear recordings in the original bit depth and sample rate. 48/24 suits me fine. Now I already know that is ridiculous pipe dream that isn't happening commercially. I hear some short recordings done by people on some of the forums like Gearslutz or taperssection or similar. Done with two or three microphones and no processing. Man that sounds so refreshing and impressive. These are usually small acoustic groups, chamber music or orchestral recitals. Too bad there is no market for those recordings. I get they aren't practical for most other kinds of modern music.

16 bit done well is enough, but just barely enough. Everyone in every part of the chain must do their job for it not to get messed up. If we can get a little breathing space even on the distribution side it probably isn't a bad thing. Only need a little though. Anything over 96/24 is just crazy, and while our bandwidth limitations are not what they used to be even that is wasteful for minor edge cases where it might be a little better.

 

[watchnerd]

In acoustic music settings, min to max dynamic range may reach 110-120 db in the presence region (there will always be a higher noise floor in the bass region). I can attest to that, from my experience of singing in a semi-professional choir which frequently performs with orchestras: Within the same piece, it can go from very very low - whispering at a level which is barely audible - to super loud. I've occasinally snuck out my iPhone with a good quality SPL meter app during practice, and I've measured peaks of 118-119 db from where I stand among the basso. This kind of dynamic range rarely gets reproduced in the home, neither by our systems nor by the recordings.

Exactly -- commercial recordings aren't allowed to leave the house with dynamic swings that wide, not even classical.

So it's a rather moot argument unless one is making and listening to one's own super dynamic recordings on a regular basis.

 

[bennetng]

These are usually small acoustic groups, chamber music or orchestral recitals. Too bad there is no market for those recordings.

Try Todd Garfinkle.

 

[Fitzcaraldo215]

It seems like a lot of hurdles need to be cleared before an audible improvement over CDs can be claimed.

In addition to the test method:

1. same definitive source master,
2. same accurate playback system,
3. well developed, consistent and accurate listening skills, Hmmm.
4. neutral listening environment,
5. control of individual subjective preferences, Hmmm.
6. etc.,
Then there are the loudspeakers.

It is probably time to remove human hearing variables and biases and just use signal comparison and analysis to do comprehensive A minus B difference detection as the first cut to look for what may or may not be audible. No matter how good double blind testing is, participants can hear what is not there.

Not all of your criteria can feasibly be met in careful scientific testing. The published scientific studies are out there, as summarized in the Joshua Reiss meta analysis paper. But, like any tests on human subjects, there is variability in the ability to discriminate and hear a difference. Some hear it, some do not, and some of the different, individual test protocols also produced a varying degree of overall results. Some tests studies showed no preference or a weak preference for hirez overall, others showed a stronger overall hirez preference. And, some testing methods were weaker than others in their scientific controls and methods. Hirez is not something that yields a slam dunk difference everybody can hear.

Even in the Harman DBT speaker testing, not all subjects, even those pre-trained as to what to listen for, prefer the speakers with the best measured performance, though overall they do beyond statistical significance. Also, I know of little drug testing that yields proven 100% effectiveness for all subjects, and the FDA approves numerous drugs with statistically significant effectiveness, but which is well short of 100%.

People vary and test methods vary. It is a fact of life. I am subjectively and anecdotally comfortable with the sonic advantages of hirez. But, others may not be.

 

[oivavoi]

Okay, these are digital only files done with Audacity. I have done such things with real gear, and the dither works up to the limits of the electronics noise of the gear.

First I created a 2 khz tone at -118 dbFS. You see here in this 1024 bin FFT what it looks like in 32 bit float.
View attachment 9750

Next we have this file saved as 16 bit without dither applied. You can see the signal disappeared as it fell below the LSB.

View attachment 9751

So now I apply dither when saving it as a 16 bit file. In this case TPDF dither. And you'll see there it is. The level shown has raised a fraction of a db as it is only a few db above the noise floor of the dithered LSB.

View attachment 9752

Here is the same file this time saved as 16 bit with shaped dither. An even lower noise floor below 5 khz in which a signal could be recovered.

View attachment 9753

Finally the same file saved as dithered 24 bit.

View attachment 9754

So you can record and recover a noisy signal at -120 db with dither. It isn't the same as the signal at 24 bit which is less noisy (though still some noise from the thermal noise limits of the gear in use). We can hear some 10-15 db into noise floors at least. So I am not sure you could hear this on a playback system set to play 0 db FS at 120 dp SPL. It might be hard to pick up in the noise using TDPF dither. You can hear such a thing somewhere around -110 dbFS perhaps. If shaped dither were used, you might hear it.

So a simple un-monkeyed with dithered 16 bit has a human perceivable 110 db or so of dynamic range. One with shaped dither might manage over most of the range more like -120 db. Finding such recordings will be difficult. It does show that if one did all work at 24 bit, and did one final downsampling with dither to 16 bit you likely are getting dynamic range equal to or exceeding nearly all playback systems.

Thanks! Very enlightening!

 

[oivavoi]

I would prefer to hear recordings in the original bit depth and sample rate. 48/24 suits me fine. Now I already know that is ridiculous pipe dream that isn't happening commercially. I hear some short recordings done by people on some of the forums like Gearslutz or taperssection or similar. Done with two or three microphones and no processing. Man that sounds so refreshing and impressive. These are usually small acoustic groups, chamber music or orchestral recitals. Too bad there is no market for those recordings. I get they aren't practical for most other kinds of modern music.

I'm in the market for those recordings. Which sections are they in on gearslutz?