[71 dB]

In everyday life the difference of loud and quiet is 30 dB, 5 bits.

 

[castleofargh]

to put a perspective, not on hearing, but on point of views, @amirm is clearly considering all the extremes in isolated conditions. it makes the practical results somehow ludicrous for music listening, yet true in the sense of noticeable to human if all the conditions are met. it's like jitter tests where people will achieve a given result for a given type of jitter in a specific test with specific signal, but then we play music instead of the test signal and the same people will stop noticing stuff a good magnitude or two sooner. it doesn't make the first test false, only very conditional and not at all representative of our practical needs for usual listening of musical content.
of course it doesn't mean that when I'm a in club (France laws require clubs to stay below 102dB(a) / 118dB(c) over 15mn), I'll get pissed off by my friend breathing very quietly 3 meters away from me and ruining the music. we all understand that hearing just doesn't work that way including @amirm . but as music could in principle be any sound at any non too painful loudness, @amirm is just covering all situations including those which will never happen. I don't think he said anywhere that we were able to notice both extremes at all time so let's not put words in his mouth.

on a practical level with my favorite songs at my desired listening levels, I have the hardest of times noticing anything at -80dB. I need instead to listen loud over a very quiet passage for such a stuff to start making a difference in my musical experience. what he says and what I say aren't contradicting each others, they simply relate to clearly different listening conditions and so we reach different results. not that controversial. I'm of the opinion that typical musical content already places us well outside of his concerns, but as the media to solve his concerns already exists(for recording and playback it's a different matter, but the media is fine), of course he should use 24bit. and I'll be sticking to 16bit for a while longer because my needs are already covered that way.

 

[Cutestudio]

And this brings up a point: This whole thread ignores the elephant-in-the-room: The CONTENT. 16bit VS 24bit is a moot point if the master is hyper-compressed and has had the top 4-8 dB brick-wall-limited off of it!

Now I do realize that that has to be done mainly at the behest of the clients(the artist, the band, or the producer or label) that want volume level 10 loudness at a volume knob setting of 2, but it belays digital audio's ability to actually sound great.

Most of us on here have probably read that on-line article about the Nirvana 'Nevermind' 24bit high-rez downloads, and what was lost. But that example proves that there is no point in a high-res deliverable if the content itself is, under the marketing guise of 'remastering', treated to a steam roller and a lawn mower set too low!

That said, 24-32bit/96-192khz sampling will continue to be the obvious choice for tracking, mixing, and post, while 16/44.1 will still be more than adequate for deliverable(CD, download, etc) for the forseeable future.

Once both artists and engineers get past this addiction to everything in a song - from the lead vocals to the rhythm section to the melody and backing vocals - being dialed up to eleven, and digital's true dynamic range is seriously exploited, then we can start talking about 'higher res' deliverable formats.

Good to read at least one other person in this thread has spotted the elephant. At last. The clipping and compression reduce the peaks by around 6dB, which can be seen with a small amount of research easily enough. Madonna's original 'True Blue' vs the remashed 'True Blue' is a case in point, compare them and you can see about 6db (the MSB) has been lost and the result is about 1/2 the peak level that was there in the first place for the same RMS level. The difference between 16 and 24 bit is an extra 256 levels below the LSB (-96dB) in 16 bit which is useful, but we have a +6dB problem with todays digital music releases.

Additionally the argument about 120dB total range is rather pointless, if you have even a simple digital level control 24 bit is the way to go even if you start by feeding 16bit in at the start. Most DACs will be what - 18 to 20bit so any replay-gain style levelling will do best output as a 24bit source. I.e. 1/16th of a 16bit signal is 12bits, from 24bits you still have 20bits. So due to further audio processing after the CD the 16bits is rather a moot point because clearly 24bit allows digital EQ (including digital speaker crossovers) and attenuation so the aim is to escape from 16bit ASAP.

Given then that 16bit is only relevant as a compact distribution medium the lack of the top bit is a serious and major problem for any HiFi. What's the distortion of a clipped or limited peak - 50%? So given we have a 50% grade distortion problem It's fascinating how we got to page 290 of a discussion about distortion levels of around 0.003%. The elephant is truly invisible.

 

[RRod]

There are genres that aren't compressed to hell. I have plenty of stuff mastered near -30dB RMS with 0dBFS peaks in my classical collection.

on a practical level with my favorite songs at my desired listening levels, I have the hardest of times noticing anything at -80dB. I need instead to listen loud over a very quiet passage for such a stuff to start making a difference in my musical experience. what he says and what I say aren't contradicting each others, they simply relate to clearly different listening conditions and so we reach different results. not that controversial.

Yes, the first thing you need is music that makes you turn the pot up high in the first place. You then need a room quiet enough to hear the errors. No real mystery, but it's simply not reality for most folks.

 

[gregorio]

on a practical level with my favorite songs at my desired listening levels, I have the hardest of times noticing anything at -80dB.

That figure would easily cover the vast majority IMO, more than 99% of people in real life listening.

And this brings up a point: This whole thread ignores the elephant-in-the-room: The CONTENT.

No it doesn't, I've mentioned the content many times in this thread and indeed my last post was largely about the content. Of course though, there's different types of content, the vast majority of music uses no more than about 8 bits, a lot of it effectively uses no more than about 6 bits but some, mainly symphonic music, can use up to 10 bits or so. This is rather a simplification though, there's a whole range of noise and noise floors to consider and many of the posts in the last few days have effectively centered around the decay/reverb tails of notes fading into these noise floors. How far/quickly they fade into the noise floor in real life (say at a performance) verses how far it's possible to fade them in mixing/mastering on a recording and then how far science tells us it's possible to hear.

[1] The clipping and compression reduce the peaks by around 6dB, which can be seen with a small amount of research easily enough. Madonna's original 'True Blue' vs the remashed 'True Blue' is a case in point, compare them and you can see about 6db (the MSB) has been lost and the result is about 1/2 the peak level that was there in the first place for the same RMS level.
[2] The difference between 16 and 24 bit is an extra 256 levels below the LSB (-96dB) in 16 bit which is useful, but we have a +6dB problem with todays digital music releases.
[3] Additionally the argument about 120dB total range is rather pointless, if you have even a simple digital level control 24 bit is the way to go even if you start by feeding 16bit in at the start. Most DACs will be what - 18 to 20bit so any replay-gain style levelling will do best output as a 24bit source. I.e. 1/16th of a 16bit signal is 12bits, from 24bits you still have 20bits. So due to further audio processing after the CD the 16bits is rather a moot point because clearly 24bit allows digital EQ (including digital speaker crossovers) and attenuation so the aim is to escape from 16bit ASAP.

1. You can't put a figure on it but typically, in the case of popular music, it would be way, way more than 6dB. Madonna's original mix would have had way more than 6dB of compression to start with, even before the original mastering, let alone after the remastering. And no, there's no way to know how much compression has been applied without access to the original recorded, unprocessed tracks. Audiophiles have got to get away from the notion that compression and distortion are bad things, on the contrary, they are essential and have been for 50 years or so. Do you really want all the guitar solos in rock and pop music to sound like a virtually inaudible series of very short duration twangs, because that's what an electric guitar sounds like without compression and distortion? And what about all the other elements/instruments in the mix? Without compression and distortion the popular/rock music you listen to would not sound anything like pop or rock music, so please stop with the "we've lost the MSB" because we've lost a whole lot more than the MSB and gained a far greater amount from that loss!!

2. Nope, even the OP discussed this. The number of quantisation values doubles for each bit of data: 16bit = 65,536 values, 17bit would therefore = 131,072 and 24bit = 16,777,216. So the difference between 16 and 24bit is obviously not 256 potential values but 16,711,280!

3. Again, no! You've made a good argument for a processing environment to be greater than 16 bits but not the distributed audio files themselves. This is why today's pro mixing/processing environments are typically 64bit float and even going back around 20 years they were 32bit float or 48bit fixed.

G

 

[bigshot]

And this brings up a point: This whole thread ignores the elephant-in-the-room: The CONTENT. 16bit VS 24bit is a moot point if the master is hyper-compressed and has had the top 4-8 dB brick-wall-limited off of it!

It isn't even an issue with dynamic recorded music. The most dynamic recorded music doesn't exceed a dynamic range of 50dB or so, and music is usually normalized to a few dB of the zero line. With redbook, that leaves a whole lot of room to play "hide the noise floor". I've done a lot of mixes where we worked at 24 bit and then bounced down to 16 at the end. We carefully compared the 24 bit mix to the 16 bit bounce down each and every time. Never a difference. That's on studio equipment with a room full of critical ears doing the checking. 16 bits is totally sufficient to cover the dynamics in any recorded music.

to put a perspective, not on hearing, but on point of views, @amirm is clearly considering all the extremes in isolated conditions. it makes the practical results somehow ludicrous for music listening

And we're in a forum to discuss how science can help improve the perceived quality of sound in recorded music. Arguing for extremes like this belongs in a forum for research scientists or perhaps studio engineers, not for home stereo enthusiasts. Demanding that a DAP have a noise floor that far down is no different than demanding that Dark Side of the Moon requires 24/192 for distribution and that frequencies above 20kHz are worth going to the mat for. That kind of obsessive pursuit of the unhearable is patently absurd.

Cutestudio, compression is a recording tool, just like EQ, reverb and levels. It isn't good or bad unto itself. It's all about how it's applied. You wouldn't like the sound of Madonna's song if absolutely no compression was applied. You'd struggle to understand the words and the subtler sound of the instruments would be plowed under. Also, I don't know what you mean by missing "top bit". Dynamic range in digital audio extends downward, not upward like burning in on analogue tape. The difference between 16 bit and 24 bit is in the quietest parts, not the loud ones. The part of the dynamic range that 16 and 24 share at the top are identical. Did you have a chance to read the article in my sig called CD Sound Is All You Need yet?

 

[amirm]

I have been listening to CDs for 3 decades and never have I heard the noise floor of 16 bit audio. All the noise I have heard is part of the recording and would be there no matter how many bits were used. I have also never heard anyone complain about hearing 16 bit noise floor.

So when you said "people" can't hear this, you were merely reflecting yourself. In that case, it would be best to say it that way.

Now as to yourself, you are not explaining any experiment that would back what you said: "People don't hear the noise floor of 16 bit audio when they listen to music, not even when the track fades away. " Have you done any conversions of 24-bit content to 16 bits while the content fades away as to know there is no difference there?

Also, how do you know what was in the recording vs what was in the channel?

Finally how did go from people can hear to don't "complain?" How would they complain without a reference otherwise?

As an acoustic engineer I have some understanding of the practical demands of dynamic range in audio and anyone who has played with 16 bit audio in a wave editor, downsampled 24 bit to 16 bit using dither understands that 16 bit is enough.

Great. Please provide an example of these where you couldn't hear the difference and let us do a listen.

 

[amirm]

Let me state my position without the back and forth as to make it more clear.

It is important in the context of audio "science" that we are true and correct to what that is. Paper after paper from luminaries in audio show that 16 bits is insufficient dynamic range (without noise shaping) to be transparent to listeners. There is also research that shows high resolution content resampled to 44.1 Khz can have audible consequences. If you are going to jump up and down and say this isn't so, please don't bother unless you have research you can put forward to the contrary. Or controlled listening tests you have performed. Otherwise, it really is useless to give me anecdotal information about what you think or hear. That is not material.

It is also true that vast majority of people and this includes audiophiles will have a heck of time telling the difference between high resolution content and CD rate. We as listeners simply don't know what these effects are and much of what we say we hear in sighted listening is not because of what we hear, but what we think we hear. Given this, why do I insist on the first paragraph above? Simple: if as objectivists we wear the cloth of science, science should not be the first thing we sacrifice to promote our message. We need to be truthful and knowledgeable about what the science says in this regard.

In my past career, I have done a ton of controlled testing and found the above to be very true. But what was also true was that I and the rest of trained listeners in my group and elsewhere in the industry could readily hear and identify artifacts that vast majority of people could not. Unless you have been exposed to this class of people (and a few gifted individuals who have these abilities without training), you can't generalize to what "people can hear."

Heck, I can teach you to hear some of the things you say are impossible to hear! I suggest not going there though as it is not good to learn to hear small differences.

So in summary, pull back a bit from extremism here. Our case doesn't hold when we go there.

Finally, all of this talk is immaterial anyway. CD as a format has had its useful life and there is no reason for us to continue to melt plastic to make it. We can deliver content online without such a constraint and vast majority of our devices already knows how to play high-res. To that end, I like to get my hands to stereo mixes prior to CD mastering. Whatever that sample rate is, I want it! If I want it at 16/44.1, I can convert it myself or download that version which usually is available anyway. I don't want my content to have been subjected to loudness compression which sadly comes with mastering the CD. By constantly defending the CD as a format, we work against this ideal. That is not right in my book.

 

[71 dB]

So when you said "people" can't hear this, you were merely reflecting yourself. In that case, it would be best to say it that way.

Yeah, but you are also hinting you can hear it. Why else would you speak for 24 bit audio if 16 bit was enough to you?

Now as to yourself, you are not explaining any experiment that would back what you said: "People don't hear the noise floor of 16 bit audio when they listen to music, not even when the track fades away. " Have you done any conversions of 24-bit content to 16 bits while the content fades away as to know there is no difference there?

Yes, I have and that's why I know I know this stuff (I knew it long before because my university studies included digital audio), but the conversion tests confirmed it for me). I studied how dither allows sounds below quantization noise floor to be heard. The noise floor must be amplified A LOT to be heard even when the signal is quieter than the noise itself!

Also, how do you know what was in the recording vs what was in the channel?

Well you can't, except there are recordings without noise to those demonstrate how 16 bit can do it.

 

[Darren G]

Again a 90 degree view...

I've owned gear that annoyed me to no end, and it was NOT because of noise.

All transducers had a sound signature, some I enjoy, some I found annoying, and (yes some of you won't like this), but something about the amp, or yes the DAC, had a repetitious quality that I found annoying.

Today I have gear that I enjoy much, but the noise level? Irrelevant. As long it is noise, random, I don't even listen to it. Sometimes it is as recorded, some times it's just noise, but noise doesn't bug me. Go to hear live music and there is massive noise, but who cares?

None of this is a sound engineer's problem to resolve, but I use an amp with tubes in the pre-amp section. I don't care if the tubes are imperfect. It is completely irrelevant to me, as a listener. What I do care about is it sounds good, and so I listen to more music. I prefer 6922 tubes because of the lower noise floor, but I am not listening to the noise anyway. I am listening to does this sound good to me?

Some gear just annoys me. My brain objects. Some gear sounds pleasant. It may well be the poorer performing gear, but again, nobody goes to a live concert and doesn't enjoy it because there is noise coming from the speakers.

Also while I have been to live performances that may reach that 120db peak, I've also left because it was so loud it was painful. Dynamic range is one variable, but it is not the only one that matters. A great master CD at 16 bits of dynamic range still kicks butt over a crap one at 24 bits.

 

[amirm]

It isn't even an issue with dynamic recorded music. The most dynamic recorded music doesn't exceed a dynamic range of 50dB or so, and music is usually normalized to a few dB of the zero line.

Where is the research and data to back this?

Heck, what is the tool that was used to determine this? Have you converted your music to less than 10 bits which this represent and failed to tell the difference?

 

[amirm]

Yeah, but you are also hinting you can hear it. Why else would you speak for 24 bit audio if 16 bit was enough to you?

As I just explained, I ask for 24 bits because that is what is used to create the music in the first place. I have no use for someone doing the conversion to 16 bits for me with who knows what scheme. Or for what reason.

But yes, I have run a public test put forward by objective blogger, Archimago and here are the results: http://archimago.blogspot.com/2014/06/24-bit-vs-16-bit-audio-test-part-i.html

foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/08/02 13:52:46

File A: C:\Users\Amir\Music\Archimago\24-bit Audio Test (Hi-Res 24-96, FLAC, 2014)\01 - Sample A - Bozza - La Voie Triomphale.flac
File B: C:\Users\Amir\Music\Archimago\24-bit Audio Test (Hi-Res 24-96, FLAC, 2014)\02 - Sample B - Bozza - La Voie Triomphale.flac

13:52:46 : Test started.
13:54:02 : 01/01 50.0%
13:54:11 : 01/02 75.0%
13:54:57 : 02/03 50.0%
13:55:08 : 03/04 31.3%
13:55:15 : 04/05 18.8%
13:55:24 : 05/06 10.9%
13:55:32 : 06/07 6.3%
13:55:38 : 07/08 3.5%
13:55:48 : 08/09 2.0%
13:56:02 : 09/10 1.1%
13:56:08 : 10/11 0.6%
13:56:28 : 11/12 0.3%
13:56:37 : 12/13 0.2%
13:56:49 : 13/14 0.1%
13:56:58 : 14/15 0.0%
13:57:05 : Test finished.

----------
Total: 14/15 (0.0%)

As you see, I achieved near perfect (14 correct answers out of 15) results. To be clear, I can't always do this. Not all content or even most of them are revealing of such differences. What is uses is what is available for download freely (from 2L site I think) and is not necessarily content that can show such differences.

And oh, both files say "24 96" but one of them is converted to 16 and then back up to 24. In other words it has no more information beyond 16 but it is presented as 24 bit as to make computer analysis very, very difficult. And to be clear, the above results are pure listening test with nothing but Foobar player playing the content and me listening on my laptop with my IEMs.

I think the files are still there. If not, let me know and I will provide them on dropbox. I would be curious to see you run them and report on whether you can or cannot hear the difference.

 

[71 dB]

Finally, all of this talk is immaterial anyway. CD as a format has had its useful life and there is no reason for us to continue to melt plastic to make it. We can deliver content online without such a constraint and vast majority of our devices already knows how to play high-res. To that end, I like to get my hands to stereo mixes prior to CD mastering. Whatever that sample rate is, I want it! If I want it at 16/44.1, I can convert it myself or download that version which usually is available anyway. I don't want my content to have been subjected to loudness compression which sadly comes with mastering the CD. By constantly defending the CD as a format, we work against this ideal. That is not right in my book.

Where do you store you downloads? You always need material to store information be it a CD disc, USB memory stick or a hard drive. Loudness war is a commercial thing. People with more sophisticated music taste buy more music which does not suffer from it. I'm listening to Joseph Schwantner's Orchestral Music on Naxos while writing this and I don't hear ANY loudness compression. If there is any, it certainly doesn't do harm. Sounds dynamic!

We defend 16 bit digital audio in this thread. Doesn't matter if it's delivered on a CD or if it's a FLAC file and anything can be mastered to sound horrible.

 

[gregorio]

It is important in the context of audio "science" that we are true and correct to what that is. Paper after paper from luminaries in audio show that 16 bits is insufficient dynamic range (without noise shaping) to be transparent to listeners.

bigshot said: ↑ It isn't even an issue with dynamic recorded music. The most dynamic recorded music doesn't exceed a dynamic range of 50dB or so, and music is usually normalized to a few dB of the zero line.

Where is the research and data to back this?

Yes, this is a sub forum of sound science but it is not a forum for ONLY sound scientists, it's a forum for audiophiles to discuss their hobby, to understand how sound works and how it applies to their hobby. The members here are listening for enjoyment to commercial music/audio releases in consumer environments, albeit typically better environments than the average consumer. This means they are not dealing with pure science, only scientific test signals, only laboratory listening environments or the ultimate limits of human hearing. They are dealing with content created by artists and engineers for normal people in normal listening environments, listening normally and for some/much of what we're dealing with there's relatively little scientific research.

For example, your question to bigshot; I've never seen scientists sit and measure the noise floor and peak levels of say a live symphony performance in the middle of the audience. Maybe there is such research but I've not seen it, have you, can you quote it? What we do have is tens of thousands of sound recordists/mixers, many decades of practical, real life knowledge/experience and the millions of albums (plus millions of films and TV shows) to which everyone here is listening. Is this knowledge/experience all utterly invalid because scientists have not confirmed it by publishing studies of real life performances? I cannot disagree with bigshot's assertion on this point, there is not the science to contradict it and it's in line with my professional experience and those with whom I've worked for more than two decades.

How is 16bit with noise-shaped dither, theoretically 120dB or so of dynamic range, not enough for any real life music recording/reproduction? And, if noise-shaped dither has not been applied to a particular master, why do you think that is?

G

 

[bigshot]

Where is the research and data to back this?

You want a scan from a book taken completely out of context or some fudged ABX test results? I'm afraid that ain't my style!

I've supervised sound mixes. I know how the system of recording and mixing works. More dynamics doesn't make for a better mix. A proper balance of dynamics does. Sound engineers want to create a mix that stays within the ear's natural dynamic range without having to take a couple of minutes to let them adjust to a new dynamic range. Human ears can hear about 45dB of dynamic range at a time. If you want to push beyond that, you have to give the ears a couple of minutes to adjust to the change. You can't just say, "I've been working at peak level and now I want to put some quiet detail at -70dB." No one will hear it. Music is recorded and mixed for human ears, not oscilloscopes and SPL meters. That means you work within the limits of human hearing to create a balance that is comfortable to listen to.

We're talking about recorded MUSIC, not abstract numbers on a page. There are aesthetic issues involved that far surpass the technical ones in relative importance. Before you raise the argument... Yes, creative decisions CAN be made in an objective manner. The biggest way that's accomplished in a recording studio is by considering the listener and creating a carefully controlled pace of contrasts to keep things interesting without being jarring or blowing out the listeners' hearing. I can tell you quite clearly that for the purposes of listening to recorded music in the home, your figures are all wrong. The simplest way to prove that is to take your favorite SACD and put it on your home stereo and turn the volume up to a peak level of 120dB. Even assuming that your living room has a dead silent noise floor and you can hear all the way down to 0dB (which it isn't and you can't), 120dB is not something you'll be able to tolerate for more than a minute or two. Go ahead and try to listen to the whole album. I bet you five bucks you can't.

Your figures do not reflect any sort of understanding of what we are talking about here... shopping for a player and speakers or headphones to listen to music in our own living room. You have some idea in your head that unless you make sure that an amp or DAC can perform properly with a test tone in an anechoic chamber (or in deep space!) it's inadequate. Sorry. That is audiophool thinking. There's a point where better numbers on the page don't result in better sound for ears. Continuing to double down on crossing every T and calculating pi out another fifty decimal places and taking into account every extreme circumstance just becomes ludicrous. You've gone so far over that line, it seems that you have no concept of what the purpose of reviewing home audio equipment is all about any more.

I've already said that I don't trust your tests. I think you enjoy skewing the truth to "prove" your absurdism ad extremum. I also don't trust your judgement. I don't think you have any sense of proportion about what is important and what is totally irrelevant. I guess what I'm saying here is that you talk and talk and nothing of any value comes out. Feel free to keep doing that if you're enjoying yourself, but don't kid yourself into thinking what you're doing is useful to anyone in any practical way.