Why do NOS dacs sound different to oversampling designs?

[abm0]

Timing is very possible at 44 khz into the picosecond range. Even within a channel. You are confusing needing bandwidth directly related to timing accuracy. Two different things. An event can be timed much more precisely than the time between samples. Transients that require 200 khz bandwidth are a different matter.

Transients that require 200 kHz are the matter I was interested in, I'm not confused about that. I don't know where you got the notion I was ever discussing the timing accuracy (the 'when') of transients rather than their speed. I was always talking about their speed and only their speed.

But OK, now I have some actual criticism to work with, I'll have to take a look at Kunchur's papers again, see if he made the mistakes you say he did.

 

[abm0]

one thing is he mistook JND's as the smallest level difference that could be heard as different. They aren't that. They are the smallest difference a person can hear and hear that there is a level difference.

Uhh, the general concept of JND is exactly that - the absolute threshold for perception of difference (ANY difference).

"In the branch of experimental psychology focused on sense, sensation, and perception, which is called psychophysics, a just-noticeable difference or JND is the amount something must be changed in order for a difference to be noticeable, detectable at least half the time (absolute threshold)."
-- Wikipedia

But of course since here he was talking about the amplitude-JND, it's going to be about level - that's where he's applying the concept of JND, like he just told you. The whole explanation was about how his experiment shows discrimination that can't be because of level-difference detection because his level differences RMS were below the JND for the frequency used, i.e. 7 kHz, which he quotes as 0.7 dB using a scientific paper by Jesteadt, Wier and Green ("Intensity discrimination as a function of frequency and sensation level.", 1977, Journal of the Acoustical Society of America). What sources are you quoting the 0.25 and 0.2 numbers from? (As a side-note, I notice that even in the main audibility thread in this forum the JND for loudness is established based on two quotes from technical/engineering books rather than from scientific studies. Not quite what I expect from an "Audio Science Review".)

 

[Blumlein 88]

Uhh, the general concept of JND is exactly that - the absolute threshold for perception of difference (ANY difference).

"In the branch of experimental psychology focused on sense, sensation, and perception, which is called psychophysics, a just-noticeable difference or JND is the amount something must be changed in order for a difference to be noticeable, detectable at least half the time (absolute threshold)."
-- Wikipedia

But of course since here he was talking about the amplitude-JND, it's going to be about level - that's where he's applying the concept of JND, like he just told you. The whole explanation was about how his experiment shows discrimination that can't be because of level-difference detection because his level differences RMS were below the JND for the frequency used, i.e. 7 kHz, which he quotes as 0.7 dB using a scientific paper by Jesteadt, Wier and Green ("Intensity discrimination as a function of frequency and sensation level.", 1977, Journal of the Acoustical Society of America). What sources are you quoting the 0.25 and 0.2 numbers from? (As a side-note, I notice that even in the main audibility thread in this forum the JND for loudness is established based on two quotes from technical/engineering books rather than from scientific studies. Not quite what I expect from an "Audio Science Review".)

Can we agree that there is evidence from testing that two identical bits of music will be heard as different at .25 db level difference? This being the reason blind listening tests are matched within .1 db because larger than that they sound different in a way that will influence the test results.

You are making the same mistake as Kunchur. I can play files .25 db different, and you'll think they are level matched. They are below the JND for loudness. But you'll think one sounds better than the other for some other reason. The reason is the level difference, but what you'll hear sounds like a quality difference. So the listeners didn't detect a level difference as level, but they heard a difference due to level all the same. It is the difference in Absolute threshold and Just Noticeable Difference. A very basic mistake for the peer review panel.

As to what the level differences were in the test signal, I forget if he noted those or described the filter parameters and it was possible to calculate them.

 

[abm0]

Can we agree that there is evidence from testing that two identical bits of music will be heard as different at .25 db level difference?

Sure, as soon as you indicate a scientific study that shows this to be the correct threshold, superseding Jesteadt, Wier and Green. For now "you're making the mistake" of asking me to believe some number just because you said it's the right one.

 

[solderdude]

The detection level may depend on what you are testing... specific frequency tones or music.

Test your own accuity here:
https://www.audiocheck.net/blindtests_level.php?lvl=0.1
you can set 0.1, 0.2, 0.5, 1.0, 3 and 6dB level differences.
This will teach you more than all the papers in the world (with test tones = not music).

Back in the days I wanted to find out myself and built this:

One can switch in the set attenuation as selected.
It taught me which levels are relevant. at both ends of the dB scale using real music in home listening conditions.

Should be obligatory for anyone arguing about levels.

Anecdotal: I could reliably tell 0.5dB apart and when 0.2dB was selected I could not reliably detect it but the sound appeared to be just 'slightly different'
Depended on the used music (did not do this with test tones as I don't listen to those)

Nowadays this would be very easy to test and blind test with music.
Use a music piece.. attenuate by 0.,1 and one file with 0.2 dB etc.
Blind tests these with ABX and you'll know and don't have to read papers.. you'll know for the only important person ... yourself.

 

[abm0]

Test your own accuity here:
https://www.audiocheck.net/blindtests_level.php?lvl=0.1
you can set 0.1, 0.2, 0.5, 1.0, 3 and 6dB level differences.

Already did, as mentioned in the audibility thresholds thread: I can detect 0.5 most of the time, though not at 100% success rate. I'm pretty sure I couldn't do better than chance at 0.2 because otherwise I would have remembered that as being my threshold. But I'll try again one night just to be sure.

Anyway, this is all beside the point if we're not testing with 7 kHz tones - it means we're not directly testing the claim under discussion from the Kunchur paper. If I'm going to spend time testing this I would rather create 7 kHz files with small loudness differences and put those in the ABX tool to see what differences I can and can't discern.

Blind tests these with ABX and you'll know and don't have to read papers.. you'll know for the only important person ... yourself.

But on this topic I'm discussing what plausible justifications (read: scientific support) might exist for promoting certain technologies (Filterless NOS DACs and hi-res recordings) to the mass market. It's a somewhat different discussion what *I* can hear and what *I* should spend money on. That's not what I'm talking about in this thread.

 

[solderdude]

The plausible explanation is very simple.

Where there is a demand people will produce and sell.
Objectively there is nothing better about it except the reproduction of illegal test signals that do not exist in any recording.

There is a market for hi-res because people want it and pay more for it.
Manufacturers all want to be able to sell gear and have the highest specs.
Chip manufacturers want to sell chips that can handle all formats people want to throw at it... and beyond.

The the explanation is simple .... money, sales, profit and its the easiest to get that from gullible people and rich folks.

the original question is incorrect.
Why do NOS dacs sound different to oversampling designs?

it should have said Why do FILTERLESS NOS dacs sound different to oversampling designs?
The answer is simple yet again.. because of the absence of a proper filter and the distortion and treble roll-off which some people prefer.
There really is nothing more to it.

I'll try to make an animation to make my point when I have time.

 

[abm0]

The answer is simple yet again.. because of the absence of a proper filter and the distortion and treble roll-off which some people prefer.

Those are the obvious explanations, both based on spectral behavior, and normally I'd accept those and just move on. But the more intriguing hypothesis to me is the "because the time-domain reproduction is better", and as long as the Kunchur papers exist I see this as a hypothesis worth entertaining. That is, until some rigorous scientific argument or other study comes along to actually properly contradict Kunchur (something I haven't seen yet anywhere).

 

[solderdude]

But time domain isn't better at all. in fact it is f'ed up as the highest frequencies are inaccurately described and can (and will) be reproduced incorrectly.
This has nothing to do with Kunchur but all with the sampling theorem and bandwidth limited systems.

A filterless DAC is not 'faster' than a filtered DAC. It just reproduces incorrectly and adds something (sharp edges and incorrect levels) that should not be there. a filterless DAC is broken by design, not better.

 

[SIY]

The Pons and Fleischman papers also still “exist” and have a few ignorami and hucksters pushing cold fusion. But they’re wrong and everyone working in science knows they’re wrong.

 

[Ron Texas]

The Pons and Fleischman papers also still “exist” and have a few ignorami and hucksters pushing cold fusion. But they’re wrong and everyone working in science knows they’re wrong.

My fridge runs on cold fusion, but the food in there keeps disappearing.

 

[abm0]

The Pons and Fleischman papers also still “exist” and have a few ignorami and hucksters pushing cold fusion. But they’re wrong and everyone working in science knows they’re wrong.

Probably because actual rigorous scientific refutations to those papers have been published and are widely known. I'm waiting to find something similar re: the Kunchur papers (but all I'm treated to so far are deep convictions, jokes and misrepresentations of what Kunchur said, i.e. Straw Man arguments).

 

[SIY]

Probably because actual rigorous scientific refutations to those papers have been published and are widely known.

That's because cold fusion is actually important. Audio nonsense isn't. There might be a published refutation out there, I haven't bothered looking. Kunchur's work was publicly gutted years ago by people like @John_Siau and @j_j who are actual specialists in audio and acoustics, so everyone shrugged and moved on.

 

[Blumlein 88]

Sure, as soon as you indicate a scientific study that shows this to be the correct threshold, superseding Jesteadt, Wier and Green. For now "you're making the mistake" of asking me to believe some number just because you said it's the right one.

First I'm aware of is part of a presentation at the AES in the 1980's by Stanley Lipschitz and John Vanderkooy. They showed that under blind listening conditions a level difference of .2 db was detected at mid frequencies in the range of our most sensitive hearing. Previously some suggested .5 db was close enough because of JND's being .7 to 1.2 db. Others thought you should use .25 db just to be sure.

Couldn't find a copy of it right at the moment.
The Great debate: subjective evaluation. from the 1981 AES journal.

 

[Blumlein 88]

Already did, as mentioned in the audibility thresholds thread: I can detect 0.5 most of the time, though not at 100% success rate. I'm pretty sure I couldn't do better than chance at 0.2 because otherwise I would have remembered that as being my threshold. But I'll try again one night just to be sure.

Anyway, this is all beside the point if we're not testing with 7 kHz tones - it means we're not directly testing the claim under discussion from the Kunchur paper. If I'm going to spend time testing this I would rather create 7 kHz files with small loudness differences and put those in the ABX tool to see what differences I can and can't discern.


But on this topic I'm discussing what plausible justifications (read: scientific support) might exist for promoting certain technologies (Filterless NOS DACs and hi-res recordings) to the mass market. It's a somewhat different discussion what *I* can hear and what *I* should spend money on. That's not what I'm talking about in this thread.

If you'll put the 7 khz tone into Foobar and test yourself you might do better than you expect. You'll probably detect it to a lower level than the one at which you can hear one is louder than the other.

As to why filterless NOS DACs exist............it is the same reason there is a multi-billion dollar industry for boner pills advertised on late night TV. Some even claim to have clinical trials in support. If people can be made to think something is going to work for more satisfaction, they'll spend money on it. The scientific support is for designs following the Shannon-Nyquist theorem about how all this really works.

The primary reason NOS DACs sound different is the frequency roll off inherent in them. Spectral tilts of .1 db/octave are audibly different. If you need citation of that I think it is in Toole's book.

 

[abm0]

If you'll put the 7 khz tone into Foobar and test yourself you might do better than you expect. You'll probably detect it to a lower level than the one at which you can hear one is louder than the other.

I can detect it pretty much every time at 0.4 dB, so I'm inclined to no longer take that 0.7 value too seriously now. And if that's wrong, it pretty much blows up Kunchur's first paper. Going to look at the other ones to see if he's done any better.

The primary reason NOS DACs sound different is the frequency roll off inherent in them. Spectral tilts of .1 db/octave are audibly different.

That may be the primary difference they make, but the same effect is also available in various warm/veiled headphones, and I'm not seeing ecstatic reviews for them like NOS DACs get, stuff like "I've never heard music before" and "I sold everything else and just kept this one, ehrmagehrd it's so guut". That's why I'm looking for alternative explanations.

 

[solderdude]

The filterless NOS R2R conversion errors are explained here in post #359 (in another thread) and also why it does not work and is conceptually flawed.

 

[j_j]

Kunchur made some fundamental signal processing mistakes, and moved on from there.

 

[j_j]

The filterless NOS R2R conversion errors are explained here in post #359 (in another thread) and also why it does not work and is conceptually flawed.

Filterless R2R means you have worse, much worse, time resolution.

 

[MC_RME]

May I stir up the discussion a bit? NOS DACs need no filter because there always exist two filters: number one is the speakers/phones that shut off around 20 kHz. Number 2 are our ears.

Of course that won't prevent me from doing it right in the first place, but....you get it.