Understanding Dither

[gregorio]

I personally pay no mind to people claiming that stuff in the least significant bits are ruining anything in any significant way.

Yes but in all fairness the article posted in the OP was not intended for consumers but for beginner mastering engineers. Mastering engineers need to consider pretty extreme circumstances, which is why mastering rooms tend to have the very best listening conditions AND also, one or a pair of the very worst speakers, both extremes. I can just about hear dither noise in my studio on some recordings but not to the extent that I can differentiate what type of dither. I sometimes spend a few minutes with the volume whacked up (in an appropriate place) and audition different dither settings but it's more of a "just in case" rather than a "this is a crucial decision".

G

 

[castleofargh]

Yes but in all fairness the article posted in the OP was not intended for consumers but for beginner mastering engineers. Mastering engineers need to consider pretty extreme circumstances, which is why mastering rooms tend to have the very best listening conditions AND also, one or a pair of the very worst speakers, both extremes. I can just about hear dither noise in my studio on some recordings but not to the extent that I can differentiate what type of dither. I sometimes spend a few minutes with the volume whacked up (in an appropriate place) and audition different dither settings but it's more of a "just in case" rather than a "this is a crucial decision".

G

sure, I really talked as me the consumer. professionals have pro needs and a work in progress isn't handled the way a final product is. I'm aware and very fine with that.

 

[amirm]

I can't hear a difference between 16/44 and 24/96 and there hasn't been any convincing controlled tests over the past 30 years that demonstrate otherwise, so I don't disagree.

There has been actually. Please see: https://secure.aes.org/forum/pubs/conventions/?ID=416

 

[RRod]

There has been actually. Please see: https://secure.aes.org/forum/pubs/conventions/?ID=416

A study crying for repetition by someone not from Meridian...

 

[bigshot]

There has never been any study that showed that super audible frequencies improved perceived sound quality in music.

 

[gregorio]

There has been actually. Please see: https://secure.aes.org/forum/pubs/conventions/?ID=416

The statement to which you responded was: "there hasn't been any convincing controlled tests over the past 30 years that demonstrate otherwise ..". Disputing this statement therefore ultimately comes down to personal opinion, whether or not one has been "convinced" by any controlled tests. Personally, I was NOT convinced by the paper you cited! 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.

G

 

[amirm]

There has never been any study that showed that super audible frequencies improved perceived sound quality in music.

The statement to which you responded was: "there hasn't been any convincing controlled tests over the past 30 years that demonstrate otherwise ..". Disputing this statement therefore ultimately comes down to personal opinion, whether or not one has been "convinced" by any controlled tests. Personally, I was NOT convinced by the paper you cited! 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.

G

Well, that is you. The Audio Engineering Society had this to say about it:

They awarded it "best peer reviewed paper."

Seemingly if we don't like the message of research, all of sudden it is bad research. Meanwhile we run with what people say online that matches our religion as gospel.

But sure, why don't you replicate the test for us that shows how easy it is to tell the difference between 16/44 and 24/96. Please post the files and double blind tests of you passing it.

 

[gregorio]

[1] Seemingly if we don't like the message of research, all of sudden it is bad research.
[2] Meanwhile we run with what people say online that matches our religion as gospel.
[3] But sure, why don't you replicate the test for us that shows how easy it is to tell the difference between 16/44 and 24/96.

1. Firstly, it's one piece of research, not the entire body of research! Secondly, I'm judging it as weak research because it is weak, nothing to do with any like or dislike of it's message. Just the title alone is dodgy (or it's a typo), it should read "atypical"!

2. By "we" I take it you mean "you"? I hope you're not speaking for me?

3. I was not talking about replicating this particular test, I was talking about manufacturing another test that would make telling the difference trivial. I don't think I could replicate that particular test, I don't think my mastering software allows an anti-alias filter with such a narrow transition band!

Look, if you want to take a single paper as absolute, proven fact, that's entirely your choice. I'm sure I could dig up a published paper or two demonstrating that tobacco doesn't cause cancer, lead in gasoline is not harmful and climate change isn't man made. I prefer to go on the body of evidence, plus professional/industry research and practice, not just one paper by someone with an agenda and a product to sell, but that's just me.

G

 

[amirm]

2. By "we" I take it you mean "you"? I hope you're not speaking for me?

Oh, I most definitely talking about you among vocal others. When someone asks me to back what I say, I always provide authoritative research so not sure why you think I quote stuff said online.

3. I was not talking about replicating this particular test, I was talking about manufacturing another test that would make telling the difference trivial. I don't think I could replicate that particular test, I don't think my mastering software allows an anti-alias filter with such a narrow transition band!

OK, please manufacture your own. How long will it take you?

 

[71 dB]

How did this topic about dither (related to usable dynamic range) become yet another debate over bandwidth and sampling frequencies?

To me it's an alarming sign of corruption within AES, that a "high-res marketing paper" by Meridian is chosen as "the best peer reviewed paper."

 

[RRod]

Well, that is you. The Audio Engineering Society had this to say about it:



They awarded it "best peer reviewed paper."

Seemingly if we don't like the message of research, all of sudden it is bad research. Meanwhile we run with what people say online that matches our religion as gospel.

But sure, why don't you replicate the test for us that shows how easy it is to tell the difference between 16/44 and 24/96. Please post the files and double blind tests of you passing it.

How did this topic about dither (related to usable dynamic range) become yet another debate over bandwidth and sampling frequencies?

To me it's an alarming sign of corruption within AES, that a "high-res marketing paper" by Meridian is chosen as "the best peer reviewed paper."

In the year they premier MQA.

 

[old tech]

Amirm

The most convincing test for me is the 2007 Myer and Moran study. It was controlled, many repetions over a lenghty period of time, had a large sample of listeners of various listening abilities (ie musicians, recording producers, audiophiles etc) and for some, using their own music, on their own stereos in their own homes in their own time.

I know, you will say but that test was ultimately flawed as it was later found that most of the SACDs were sourced from 16/44 masters. However, that was why that test was so convincing. You see, over many years all those SACDs were released in markets around the globe yet no-one noticed, not one audiophile listenter, not one golden eared reviewer, not even one subjective reviewer. Most were convinced that their SACDs sounded superior to CDs. In the end, the lack of provenance was confirmed by testing, not listening.

I know this off topic to dither, but I think I can take liberties being the OP!

 

[RRod]

I know this off topic to dither, but I think I can take liberties being the OP!

We're not too far off, considering the gold-star paper mentioned above made pains to use a type of dither that no one uses...

 

[bigshot]

There has never been any study that showed that super audible frequencies improved perceived sound quality in music.

I'm going to quote myself because it's important to look at the purposes we use digital audio for... listening to recorded music in the home. There may be some obscure situation or some specific sort of signal that a study can try to create to "break" the perfection of good old redbook. But that doesn't change the fact that for the purposes of listening to recorded music in the home, redbook is already overkill.

It takes some convoluted logic to claim that because a format is capable of reproducing super audible sound, that super audible content is important to listeners. That is a self contradiction. Instead of looking to the AES, we should be looking to audiologists and medical researchers to prove where the limits of human hearing lie. They would know better than people who design audio equipment. I have actually done that. The range of human hearing goes from 20Hz to 20kHz. And that is the broadest range any person can hear. Most people don't even hear that much. If I remember correctly, the world's record for the highest frequency ever heard by human ears was 23kHz. That is less than one note on the musical scale above 20kHz. And up at the top of the range of human hearing, and at the lowest point down at the bottom, human ears can't perceive pitch, only sound pressure.

The extreme ends of human hearing flat out don't matter for reproducing music. Audiophiles routinely listen to music without the top half octave and bottom half octave and still call it an audiophile recording. Why would a frequency two octaves above that have any impact at all on the sound quality of music? Answer = it doesn't. Superaudible frequencies are as useless to music as teats are to a bull hog. All they do is make playback less convenient and more expensive. It's a waste of time and money.

The same goes for noise floors. In one of the links in my sig, Ethan Winer takes the most annoying buzzer sound possible, and mixes it in at different levels under music. By the time the buzzer gets to -45dB, it has completely disappeared at normal listening levels. Redbook is capable of noise floors many many times beyond that. Why do we need 24 bit playback in our home? The answer is, we don't.

All of this focus on the extremes serves a purpose... the purpose is to convince people with OCD and very little understanding of how audio technology works to spend more money on their equipment than they need to. I can't think of any other reason.

 

[amirm]

The most convincing test for me is the 2007 Myer and Moran study. It was controlled, many repetions over a lenghty period of time, had a large sample of listeners of various listening abilities (ie musicians, recording producers, audiophiles etc) and for some, using their own music, on their own stereos in their own homes in their own time.

I know, you will say but that test was ultimately flawed as it was later found that most of the SACDs were sourced from 16/44 masters. However, that was why that test was so convincing. You see, over many years all those SACDs were released in markets around the globe yet no-one noticed, not one audiophile listenter, not one golden eared reviewer, not even one subjective reviewer. Most were convinced that their SACDs sounded superior to CDs.

Unfortunately that list also included Meyer and Moran themselves! Here it is from their paper:

And I distinctly remember reading about of use of CD masters for the SACD layer in audio magazine so it was present then for sure.

Regardless, in this day and age the point is moot: I like to get the original stereo master as created. I have no need for anyone in the middle to convert it to 44.1 Kh, 16 bits or whatever including their theory of whether dither is or is not needed.