[bigshot]

Multichannel sound, digital equalizers and DSPs are the only electronic/digital improvement in sound I've noted in the past couple of decades. Every digital audio product I've had since I got my Macintosh 8500AV in 1995 has been audibly transparent. I'm sure there are bad ones out there, but on the whole, sound quality is pretty much perfect.

 

[Don Hills]

[QUOTE="csglinux, post: 13848731, member: 399084"... But, I've often wondered if - one day, with the perfect recording - how nice it would be to crank the volume all the way up in a very quiet/silent passage and still not hear any noise floor.[/QUOTE]

You must be a fan of John Cage's music:
https://en.wikipedia.org/wiki/4′33″

 

[amirm]

while I agree that simply adding the noise of the room to the music's dynamic is an oversimplified idea. countering it with audibility measured for white noise where the room's own noise floor is now the loudest signal, that's IMO even more inadequate to depict realistic listening or listening needs.
sure enough noise isn't some opaque paint spread over the stuff below, but masking is still very much a reality, so is the human's limited instantaneous dynamic range.

Masking is content dependent. When it is not there, then the noise floor is there to be heard.

What we are trying to do is define a channel which can be shown to be error-free. Once there, the system will maintain its performance regardless of what content you play.

Notion of masking is very useful when we are looking at a signal and noise/distortion around it. Such is not the case here. I can have low frequency music notes, and mid to high frequency noise that would be audible with it due to lack of masking.

As an analogy, we can say that the restaurant doesn't need to wash their dishes if we limit ourselves to food that is able to kill the bacteria in it. That restriction would not fly in that situation, and as such, doesn't work here either.

 

[JaeYoon]

Masking is content dependent. When it is not there, then the noise floor is there to be heard.

What we are trying to do is define a channel which can be shown to be error-free. Once there, the system will maintain its performance regardless of what content you play.

Notion of masking is very useful when we are looking at a signal and noise/distortion around it. Such is not the case here. I can have low frequency music notes, and mid to high frequency noise that would be audible with it due to lack of masking.

As an analogy, we can say that the restaurant doesn't need to wash their dishes if we limit ourselves to food that is able to kill the bacteria in it. That restriction would not fly in that situation, and as such, doesn't work here either.

I'm not sure about that analogy.
One could potentially cause bodily harm to patrons that would consume that food.

Someone listening to masked frequencies would be in no chance of same potential of danger of sickness.

 

[bigshot]

What we are trying to do is define a channel which can be shown to be error-free. Once there, the system will maintain its performance regardless of what content you play.

If you define a system that outperforms the thresholds of human hearing, they don’t have to be error free.

A better analogy would be... if you ditch classes at school to go fishing and your mother doesn't find out, you don't get in trouble. What she doesn't know can't hurt you! That's masking in a nutshell.

 

[Darren G]

A question/thought for the group here (and a 90 degree thought)...

At one point in time I was working on a MPEG-1 decoder. There was limited content available to test with at that point in time, but as someone starring at the output of decoding every day, what I became aware of is macro blocks, and motion artifacts. I would show this to others and their reaction was very different. In their minds it was so clear, and 'perfect', keeping in mind they were use to seeing video played back from films, or tapes, where noise was the norm.

What I realized, or at least started to question was, our brains are very good pattern matching machines. What bothered me about MPEG-1 wasn't the errors so much as the repetitive nature of those errors. Eventually that is all my brain locked in on.

Okay so the leading question then is this...

Measuring noise is only part of the story. The more difficult measurement is repetition of noise. See I am thinking our brains are very good at masking noise, but if the noise is repetitious, that is what our brains lock on.

If we ask the question is any bit of noise below audible level, yes it is inaudible, but if there is a pattern to the noise, do our brains pick up on that, and either find it annoying or pleasant?

--

Just some thoughts to throw into the mix...

 

[71 dB]

The coding artifacts of MPEG-1 video are of course perceptually gigantic compared to quantization errors of 16 bit audio.

 

[Darren G]

Agree 71 dB. We are well into the area of diminishing returns. Still, my question is subtle... and I'll use analogy to explain what I am thinking (it is not perfect for sure).

Imagine a clear window, it is near perfect, but there is a speck of dust.

Many of us ignore the speck of dust, but what we have no clarity (or clearly definable measurement for), is do some people's brains focus almost entirely on that speck of dust, while others just see the clear window?

It's our human norm to believe our perception is universal, but we have many examples of rare individuals who excel in certain areas because their brains are wired differently.

I am not suggesting the music industry focus on the rare .000(add 0's here)1 percent, only that I do think it is a fair question.

 

[bigshot]

I would say that inaudible is inaudible no matter what the pattern. There certainly are ways that noise can be more annoying, but if it's below the threshold, it's like the tree falling in the forest with no one there to hear it.

 

[castleofargh]

Agree 71 dB. We are well into the area of diminishing returns. Still, my question is subtle... and I'll use analogy to explain what I am thinking (it is not perfect for sure).

Imagine a clear window, it is near perfect, but there is a speck of dust.

Many of us ignore the speck of dust, but what we have no clarity (or clearly definable measurement for), is do some people's brains focus almost entirely on that speck of dust, while others just see the clear window?

It's our human norm to believe our perception is universal, but we have many examples of rare individuals who excel in certain areas because their brains are wired differently.

I am not suggesting the music industry focus on the rare .000(add 0's here)1 percent, only that I do think it is a fair question.

when the little bunny poops in the woods, on the mountain I see from my window maybe 10km away. I don't think the change in panorama is something I should care about. I know that this fracking rabbit is taking a crap on my panorama probably several times a day, I can be mad about the reality of it and if I go walk in the mountain, I could end up finding poop, somehow proving how right I was to be mad about it. I can convince myself that because it is measurable, and even noticeable in a totally different context, then it matters. I can even feel like the beautiful scenery now sucks because of the rabbit. I can buy traps and catch the rabbit, and it would probably make me feel like the view is better than ever.
there is no limit or rule to what someone must consider significant and how much mess is going on in his own mind. but there is absolutely a limit to what a human eye can see when looking at my window. and the rabbit on the mountain on the other side isn't one of them.

to be clear there is nothing wrong with going highres, or trying to do even better than what's necessary or relevant. the simple idea of going beyond a limit is in itself a strong motivation for many people. but humans have thresholds and those thresholds are determined through practical testing, not through wishful thinking, sighted tests, and made up hypotheses. personally I am not against highres, I'm against pushing highres for BS reasons.

 

[Cutestudio]

The issues with arguing that all noise is the same and that statistical methods can make 16bit audio as accurate as 24bit remain as spurious as ever. This 16vs24 bit subject however is a discussion in theory only as in the real world we have a very limited choice of most music, generally between mangled mp3, mangled CD spec, not-so-mangled Mastered-For-iTunes, or mangled 'hi-res' versions sold by various chancers. Vinyl is available for some sure, some is just a 3rd rate copy of the mangled CD spec, but some is better mastered. Norah Jones's 'The Fall' is totally mashed on CD - it makes the Chilli Peppers Californication seems quite listenable - but the vinyl is quite good. Or it would be if they'd used some decent vinyl instead of recycled sunbeds and coffee cups to press it on.

I've talked with people who have bought (quite a lot of) hi-res audio and then found it's just the same as the regular 44.1/16 when viewed in a waveform editor like Audacity. You can't see the fine detail but you can see the usual brick shaped waveform with the usual damage as the dynamics were rolled out of it by a mastering 'engineer' so everything turns into a nice dull flat noise to be turned down and ignored by the sucker who's bought it..

Genuine hires can have the advantage of a frequency response. The dull 'you can't hear above 20kHz' mantra may have been fine in the 1930s, but since stereo was invented you can have all types of beat frequencies interacting between the channels, so there's a ready explanation of why a higher HF limit may sound better. The luddites would never wear a distribution format upgrade though.

However still the main problem with both hires and the regular 44.1/16 is the lack of the MSB (due to the insidious and endemic loudness war), so you end up with compression/clipping forming a series of flats in the waveform that puts many a mediaeval castle's castellations to shame. People who don't realise that the MSB is missing don't spend their days looking at the waveforms, it's not my opinion: it's a simple fact for most modern music today. The people who 'know' this isn't true need not reply, merely look at their own tracks in a waveform editor to see for themselves.

There is an argument I've heard that the loudness war is partially caused by 16bit recordings sounding a little lifeless when not fully driven but I think there are a number of factors that have left us in the situation today whereby we can argue about 16bit vs 24bit for years: but still not have the opportunity to buy a decent unclipped/compressed recording of much at all in either format.

24 bit does have some tangible benefits to digital audio however, even just as an output format: one of which is the ability to use a digital level control without the degradation of signal that a 16bit output exhibits. You can demonstrate this by simply turning down the digital level as you turn up the analog level. The degradation of the 16bit signal is quite pronounced but the 24bit is far more usable as a level control without impacting quality too greatly.

 

[bigshot]

Hey Cutestudio! I would chop up your post and answer it line by line, but that would be annoying and would just make you mad. It's a lot nicer just to point out that if you're interested in this subject, you should read the article and watch the two videos in my sig file. The videos have downloadable audio files associated with them so you can hear the truth for yourself. Obviously, you've spent some time thinking about how digital audio works, but I think you'll be surprised at how much more there is to learn than what you already know. It's always good to learn, right?

Start with the article, and after you have a chance to digest it, then watch the videos. Sound Science is fun!

 

[Darren G]

when the little bunny poops in the woods, ...

LOL, what a great analogy. I am objective (prove it) type, so you'll always find me on the side of the mountain that doubts anyone is hearing a gnat fart a mile away.

On the flip side, it is quite interesting how quantization noise is something some human brains may pick up on, and the solution is as simple as adding some random noise. I am over extending my knowledge here so sorry in advance for any misinformation.

What I am trying to understand is vague anyway, but the gist of it is, is there a threshold range of noise that is essentially inaudible, but if it is repetitious in nature, that (for some) elevates what should be inaudible into the perceptual realm of audible (conscious or not) for some?

 

[amirm]

If you define a system that outperforms the thresholds of human hearing, they don’t have to be error free.

And that is exactly what we are doing. We measure the system level of noise, compare it to threshold of hearing, then determine the loudest real life music we can find, and the required dynamic range falls out of that. From Fielder paper's summary (which is also in my article):

Using the simple math of 6 db/bit we see that we need 122/6 = 20 bits. 16 bit audio at 96 dB will not do it.

We can use noise shaping to improve 16 bit dynamic range in audible band but then we run into the limited bandwidth of 44.1 Khz. And second real life problem that you have no assurance that when the music was produced, noise shaping was used.

This is what I call error free channel. It is perceptually so. An objective error free channel of course does not exist when it involved analog signals we hear.

 

[bigshot]

I've said it before and I'll say it again. If you aren't listening to your music at 122dB, you don't need a dynamic range of 122dB. And the noise floor of a typical living room is between 35 and 40dB, not 20 to 30. Your numbers are wrong. If you get out a SPL meter and check for yourself what those numbers represent in real world sound, you'll see I'm right.