Understanding Dither

[71 dB]

Did you read what I wrote? It had nothing to do with the points you are making. Regardless, how do you know how much resolution those masters have? Do you even know how to compute that? If not, then I don't want you or anyone else in charge of truncating it to that amount. I can play the mastered format on my system thank you very much. Keep your hands away from my bits!

I read that you want original high res studio masters without anyone downsampling them to 16/44.1. Real life ADCs have dynamic range of about 20 bits and those bits aren't used optimally (that's the point of having extra bits in music production), so maybe there's 18 bits of "real resolution", but even that doesn't make it better than 16 bit, because nobody hears the difference in music listening. You'd have to listen to the loud parts in the evening and the go to sleep and in the morning listen to the silent parts and even then it would be a huge challenge. Totally impractical.

Nobody is touching your bits so please calm down. All we are trying to do is to convince you and others that high res digital audio doesn't offer anything but bigger usage of storage space. What comes to being able to "compute" things related to digital audio: Master's degree in electric engineering (majoring acoustics and signal prosessing) should help a lot with that.

Adding bits to increase dynamic range seems tempting, but every new bit is only half of the previous smallist one in significance (literally!!) so the relevance of adding bits dies away very quickly! In music listening 8 bits is not enough at all, 12 bits starts to be enough, 16 bits is overkill, 20 bits is OVERKILL and 24 bits is O V E R K I L L !!!! In your 24 bit music files the upper 14 bits are relevant and the lower 10 bits are waste of storage space.

 

[71 dB]

Which music? Let's see the clip names and how you determined their bit depths.

Of course electronic music can be produced having dynamic range of say 140 dB, but human hearing does not hear such dynamic range at the same time! Loud sounds raises the threshold of hearing. Any practical music uses much smaller dynamic range and we need dynamic range so that in silent parts we don't hear the noise floor.

I am not seeing critical thinking when a paper that has absolutely nothing to do with MQA, is criticized on basis of its principals also inventing MQA. It is just an emotional rant that has no place in a professional discussion about the topic at hand.

Meridian has lost my trust with their MQA so I am suspicious of anything related to that. MQA has to do with high res and this paper is about hi res, isn't it?

 

[bigshot]

Of course electronic music can be produced having dynamic range of say 140 dB, but human hearing does not hear such dynamic range at the same time!

It's like saying someone wrote the entire Declaration of Independence on a grain of rice once for the Guiness Book of World Records, so therefore it's impossible to properly read the Declaration of Independence without a microscope. We can only deal with one level of scale at a time. Listening to music with a 120dB dynamic range would be like reading a book where every word was a different type size, varying from 3pts up to 120pts.

 

[RRod]

Examples were already given in the 24 vs 16bit thread of reverb tails of music recorded at extremely low levels in a modern concert hall. There's nothing there to hear between 24-bit and 16-bit shaped. Might 16-bit rectangular have been audible? Maybe, but that's not best practice is it? And even if it were audible, it would be so for a whole final second of the track. Hell, I've seen plenty of tracks where they cut off the reverb early anyway.

In the Boston test, in a -19dbA room, the listening level had to be explicitly raised to hear the noise added by a 16-bit ADC/DAC conversion, and then it was only on a disc that has about a -35dBFS RMS level. Here we're avoiding the extra A in the chain and can use optimal dither; what's left to hear? This is why I find the paper in question so fishy. I would LOVE to have their test files.

 

[amirm]

RECORDED music. You can stick a SPL meter in the bell of a trumpet and record huge peaks, but show me a recording that has much over 60dB over the noise floor.

SPL meters are psychoacoustically blind. You ear does NOT hear like a SPL meter, far from it. Any such analysis needs to take into account the huge non-linearity of your ear with respect to frequency.

Look at the one labeled "threshold" and how it varies extensively in SPL depending on frequency. 60 SPL at 20 Hz would be dead silence, but near 70 dB over threshold of hearing at 3,000 Hz!

As I said, until you read and understand the science instead of repeating stuff you read online, you will have a hugely mistaken idea of what any of these discussions are about.

 

[amirm]

Examples were already given in the 24 vs 16bit thread of reverb tails of music recorded at extremely low levels in a modern concert hall. There's nothing there to hear between 24-bit and 16-bit shaped.

Shaped? What percentage of your music has shaped dither? You don't know, right?

Such shaped dither can also overload the ultrasonic range and have undesirable consequences for your audio gear oscillating, creating intermodulation distortion, etc. Now if you have higher sample rate, then you have plenty of space to hide them and not worry.

Still, all of this is academic driven by fear of high-resolution music for no good reason whatsoever. It is like campaigning that people should not have too clean of water.

 

[amirm]

Of course electronic music can be produced having dynamic range of say 140 dB, but human hearing does not hear such dynamic range at the same time! Loud sounds raises the threshold of hearing. Any practical music uses much smaller dynamic range and we need dynamic range so that in silent parts we don't hear the noise floor.

That is not the answer to the question I asked. Indeed it has nothing to do with what I asked.

This is what you said again:

Full dynamic range of human hearing is even more than "20 bits", but we don't need it, because music uses much less.

I asked you to show the clip names in your library and their true bit depth. Do you have any such data or should we conclude properly that this is a made up notion and no assessment was made by you or anyone else who keeps saying this?

 

[amirm]

I read that you want original high res studio masters without anyone downsampling them to 16/44.1. Real life ADCs have dynamic range of about 20 bits and those bits aren't used optimally (that's the point of having extra bits in music production), so maybe there's 18 bits of "real resolution", but even that doesn't make it better than 16 bit, because nobody hears the difference in music listening.

Well, we were told it was actually easy to show such difference:

Or to be more precise, it convinced me of something I'd already been convinced of (15 years or so before the paper was published), namely: That it's entirely possible to manufacture an abnormal condition or set of conditions under which it's trivially easy to differentiate 16/44 from 24/96.

Beyond that, you are going by your limited knowledge of who has heard what. What types of controlled testing have you done to make such general assessment? Are you going by your own ears or someone else's?

Archimago put forward a nice test of 24 bit versus 16 with excellent countermeasures to deter mechanical analysis of its bit depth. Have you run it? I have, and here is the outcome for one of the clips in double blind ABX test:

Here is the original link: http://archimago.blogspot.ca/2014/06/24-bit-vs-16-bit-audio-test-part-i.html

Let's have you all run the same test so that we can determine if we are operating from the same point of critical listening ability. 'Cause it sure seems like we are generalizing from our hearing to what everyone else is hearing (or worse, not ever experimenting).

 

[bigshot]

SPL meters are psychoacoustically blind.

OK. Then stick your ear in the bell of a trumpet and listen to a full forte blast and you will hear something up around 120dB. But you aren't ever going to get that in recorded music. Music is performed for comfortable listening levels, not deafening ones.

I've corrected my original post.

Anyone can cheat a bitrate test if they grab bits of the track at the end and turn the volume up to deafening volumes, right? right? (crickets)

Water can't be too pure for our precious bodily fluids!

 

[RRod]

Shaped? What percentage of your music has shaped dither? You don't know, right?

Such shaped dither can also overload the ultrasonic range and have undesirable consequences for your audio gear oscillating, creating intermodulation distortion, etc. Now if you have higher sample rate, then you have plenty of space to hide them and not worry.

Still, all of this is academic driven by fear of high-resolution music for no good reason whatsoever. It is like campaigning that people should not have too clean of water.

Actually a lot once you get past a certain vintage, having listened to most of my stuff with a spectrogram going at some point. This is in classical, though, which is arguable teh genre were someone might actually hear something, so I assume they are more principled about it. And if shaped dither were causing distortion artifacts, I haven't managed to hear them on even a bright HP like the HD800 with the volume cranked for Mahler. How much audible distortion would we even expect from the addition of a signal at ~ -90dBFS?

I agree with the assertion, actually, that there's no need to have 16/44.1 delivery any more. With modern lossy codecs handling 32-bit FP there's no reason to worry about a final dithering; just release a hi-res version to buy at CD-ish prices and use something like Opus for streaming.

 

[amirm]

Anyone can cheat a bitrate test if they grab bits of the track at the end and turn the volume up to deafening volumes, right? right? (crickets)

Well, then post one of these outcomes if anyone can do it. I provided a link for one already. When can we see the results?

 

[71 dB]

That is not the answer to the question I asked. Indeed it has nothing to do with what I asked.

This is what you said again:


I asked you to show the clip names in your library and their true bit depth. Do you have any such data or should we conclude properly that this is a made up notion and no assessment was made by you or anyone else who keeps saying this?

Sorry, I think what you ask is irrelevant. It doesn't matter what the bit depth is. What matters is how large dynamic range you can hear. Threshold is NOT -10 dB at 3000 Hz when you listen to insanely loud music!! It's much much much higher. This is the whole point and it seems you just don't get it.

 

[bigshot]

Well, then post one of these outcomes if anyone can do it. I provided a link for one already.

Yes. You are correct. When you arrived here at Head-Fi you posted logs for a blind test where you cranked the volume on a tiny bit of sound and claimed that you had "trained" yourself to hear noise floors.

 

[amirm]

Yes. You are correct. When you arrived here at Head-Fi you posted logs for a blind test where you cranked the volume on a tiny bit of sound and claimed that you had "trained" yourself to hear noise floors.

That was your accusation which you continue to unprofessionally make at every chance. Just because you think you have to cheat in your exam to get a good grade, doesn't mean others do the same.

BTW, you are welcome to play any trick you want to pass such a test. But seamingly are incapable of doing so despite infinite amount of free time to keep posting on this forum.

 

[old tech]

Shaped? What percentage of your music has shaped dither? You don't know, right?

Such shaped dither can also overload the ultrasonic range and have undesirable consequences for your audio gear oscillating, creating intermodulation distortion, etc. Now if you have higher sample rate, then you have plenty of space to hide them and not worry.

Still, all of this is academic driven by fear of high-resolution music for no good reason whatsoever. It is like campaigning that people should not have too clean of water.

What fear of hi res music? I do listen and purchase hi res music if that is the best mastered version of an album. The only fear I can read on this thread is yours of 16bits, or the inexplicit fear of going from 24bits to 16.