GoldenSounds passes apparently ABX test for DACs (NOT Really)

[Blockader]

According to the delta wave delta plots, he highest peak around 20khz was -46db. If he was listening at 110db with his headphones, that would make the peak at 64db. Due to how equal loudness contours work, that's going to be audible like a peak around 52db at 2000hz. 52db is totally audible especially in the last octave when there are no peaks to mask it. (even 20db is audible)

In other words, what Golden Sound did isn't extraordinary. Anyone who is trained and can hear frequencies between 20-21 kHz can pass the test with the right loudness levels.

 

[kemmler3D]

by this definition, a device will never be transparent since you can always construct a situation where it is not.

Well, sure, if you overload the inputs or something, anything will stop being transparent. Maybe I should have qualified that by saying "without abusing the equipment". Obviously every real-world piece of equipment has a limited range of operation.

I just think this current case is interesting because if you had asked me "Can someone tell two reconstruction filters apart if there's only an appreciable difference above 20khz" I would have said "very unlikely, almost impossible", but here we are. This is the kind of surprising event I like.

Is this relevant to anyone's music listening enjoyment? Only in the "overkill soothes my upgradeitis" sense.

 

[SIY]

If he was listening at 110db with his headphones, that would make the peak at 64db. Due to how equal loudness contours work, that's going to be audible like a peak around 40db at 2000hz.

Is that valid in the presence of 110dB sound?

 

[oleg87]

In other words, what Golden Sound did isn't extraordinary. Anyone who is trained and can hear frequencies between 20-21 kHz can pass the test with the right loudness levels.

I don't find it unbelievable that someone somewhere along the distribution of human perceptual ability may hear this difference, but I don't think "anyone" can.
I think I've seen one person report being able to (unreliably) repeat this feat in this thread. I think I've got a pretty decent ear (though I certainly can't hear 20khz) but I couldn't find a tell playing it at 0.5x rate even at volumes I found quite uncomfortable. Maybe if I spent more time trying to hear something, but life's too short.

 

[dasdoing]

Maybe I should have qualified that by saying "without abusing the equipment". Obviously every real-world piece of equipment has a limited range of operation.

But that's not what I mean. boosting volume up "500%" on a looped quiet passage has been mentioned. you are not abusing anything, but are creating a situation that will never happen in real life.

 

[kemmler3D]

will never happen in real life.

Well, maybe not just listening to music, but in professional use "looping a short passage at elevated volume" is a very realistic use case...

I'm just feeling like this whole thing is a tempest in a teapot. The guy apparently passed an ABX using cues near ultrasound. This is interesting (to me) but it's not a stop the presses moment, either. It doesn't challenge any accepted truths about audio or hearing that I'm aware of. So I don't see the need to move goalposts and say it doesn't count because the use case isn't realistic enough. Even if you take it at face value it doesn't really change anything, so I'm not sure why it's even worth arguing over.

I think the questioning of method is quite valid and useful, we want to know exactly how good the test was of course. But whether the test is informative or meaningful... if you don't think so, nobody's forcing anyone to read the thread, and I'm not aware any controversial follow-up claims have been made?

 

[danadam]

that would make the peak at 64db. Due to how equal loudness contours work, that's going to be audible like a peak around 52db at 2000hz.

How do you know what the curves look like past 13/14 kHz?

 

[board]

One the other hand generalizations don't work either for us.

If we get under the umbrella of "most people" we all (at least the ones who do actual tests) suspect that a 40dB SINAD DAC (dominated by low order distortion and not noise) would be indistinguishable by a 110dB SINAD one for most people out the street* who casually listen whatever music they listen to.

So?We would draw a line there and say is adequate because most people (the absolute vast majority) don't hear or don't care to hear the difference?
We care about engineering excellence (and I wish we all cared the same about safety,reliability,etc as is the insurance companies joyride to refuse claims based on device's certifications) even if it does not translate to better SQ for most people (and over some not scientifically defined thresholds,for all people) .

Switching between those two sometimes helps the narrative,but that's not the scientific way.

*or not so out the street but by conscious listeners too,read a little up this post on the thread here

Yep, I agree. I think most of us agree that this hobby is reserved for a tiny percent of the population who actually cares about these things. I have given ABX tests to several girlfriends and even my aunt, who is diagnosed with something akin to PTSD and therefore was very nervous about the test, and they all passed, but I honestly don't think many of them really cared about the fact that they could actually hear a difference and that there were also tests where I could pass but they couldn't, nor do I think they would want to listen to music that way - to them it's melody, rhythms, lyrics, not tiny differences in timbre, brightness, murkiness, etc.
I think most of us on this forum also agree that some people can pass certain tests while others can't, and some people care about it while others don't, but it becomes problematic when people can't pass a test yet insist they can actually hear a difference.
I've seen people claim they could hear a difference, and when they failed a test they accepted that it had all been imaginary, and that's the kind of people I like, but too many people insist the test must be broken if they fail - which is often like religious people or people who believe in zodiacs, tarrot cards, conspiracy theories, etc.
I do also find it a bit silly to buy a certain product "because in this specific song this split second sounded different". I think we should buy something because it makes the majority of our music sound (much) better, not a tiny difference in very specific situations.

 

[VientoB]

According to the delta wave delta plots, he highest peak around 20khz was -46db. If he was listening at 110db with his headphones, that would make the peak at 64db. Due to how equal loudness contours work, that's going to be audible like a peak around 52db at 2000hz. 52db is totally audible especially in the last octave when there are no peaks to mask it. (even 20db is audible)

In other words, what Golden Sound did isn't extraordinary. Anyone who is trained and can hear frequencies between 20-21 kHz can pass the test with the right loudness levels.

We don’t know how loud he was listening at. Let’s wait and see.

 

[Westsounds]

He has Golden Balls, what did you expect!

 

[RandomEar]

According to the delta wave delta plots, he highest peak around 20khz was -46db. If he was listening at 110db with his headphones, that would make the peak at 64db. Due to how equal loudness contours work, that's going to be audible like a peak around 52db at 2000hz. 52db is totally audible especially in the last octave when there are no peaks to mask it. (even 20db is audible)

In other words, what Golden Sound did isn't extraordinary. Anyone who is trained and can hear frequencies between 20-21 kHz can pass the test with the right loudness levels.

I also tried to guesstimate the level and the threshold, but there's a lot of unknowns. To state that it's "totally audible" is just plain wrong. To name a few caveats:

• We don't know the exact SPL the test was performed at. Maybe it was 95 dB, maybe it was 110. Makes a big difference.
• We don't know if any EQ was used.
• The ISO 226 equal loudness contours are only defined up to about 14 kHz, everything above that is estimated and - at 21 kHz - highly dependent on the individual. Most people won't hear anything at 21 kHz, regardless of the SPL.
• Cameron appears to be using a Hifiman Susvara. While headphone FR measurements become unreliable above 8 to 10 kHz, the level at 21 kHz will be lower than the level at 2 kHz. From the review, the difference appears to be >10 dB. This difference adds to the (unknown) one from the equal loudness contours.

I think it's fun to discuss and analyze this case. But at the moment, we still lack the necessary data to draw any reliable conclusion.

 

[Sokel]

I also tried to guesstimate the level and the threshold, but there's a lot of unknowns. To state that it's "totally audible" is just plain wrong. To name a few caveats:

• We don't know the exact SPL the test was performed at. Maybe it was 95 dB, maybe it was 110. Makes a big difference.
• We don't know if any EQ was used.
• The ISO 226 equal loudness contours are only defined up to about 14 kHz, everything above that is estimated and - at 21 kHz - highly dependent on the individual. Most people won't hear anything at 21 kHz, regardless of the SPL.
• Cameron appears to be using a Hifiman Susvara. While headphone FR measurements become unreliable above 8 to 10 kHz, the level at 21 kHz will be lower than the level at 2 kHz. From the review, the difference appears to be >10 dB. This difference adds to the (unknown) one from the equal loudness contours.

I think it's fun to discuss and analyze this case. But at the moment, we still lack the necessary data to draw any reliable conclusion.

We do know about EQ,at least at the foobar's side there's none,it would be reported at the ABX log.

Example:

 

[SuicideSquid]

I think this is where we disagree. I want to know if gear is transparent in all cases, or just most cases. Just for the sake of knowing, not because it actually matters in real life.


I know this of course, but I think it's interesting anyway.

It's not like I'm going to choose gear on this basis or think about it day to day. I just think it's interesting to explore what's possible and what's not.


Yes, but that's not the point of talking about audio science on the internet. The point of that is to have interesting discussions about audio, and I happen to find this one interesting.

If you want to say I shouldn't find it interesting, well, sorry I guess?

Interesting in the abstract? Maybe - personally I don't find much interesting in someone being able to distinguish differences between audio settings in a contrived test that has no relationship to actually uses, but if you do, that's fine. But is this useful or meaningful information? No.

 

[Head_Unit]

What gets muddled and lost in that - as it so often does in these discussions - is that if you can reliably detect a difference...that says exactly nothing about...preference,

Hurrah! I remember one demo at where the dealer replaced a big expensive speaker cable with an insanely expensive bigger cable.* There was definitely a small difference in the violin sound, repeatable. But both sounded fine, just different. Forward in time we swapped a Denon AVR-X3600H for an Anthem AVM 70 + ATI 525NC. A difference in the...3D shape? of the image. But for IIRC $6000 more it was NOT better, just different. Probably a difference in the Audyssey XT32 versus ARC Genesis room corrections.

*It's too long ago to remember model numbers, though I do know the speakers were Apogee Scintillas driven by a Mark Levinson very expensive amp, expensive CD player (also Levinson? Did they make one?) playing "In The Hall Of The Mountain King." The 2nd cables were MIT with a big network attached, doubtless interacting a lot with the Scintilla's 1 ohm impedance. Therefore from an engineering point of view not a shock to hear a difference. With normal speakers I've never noticed a difference changing cables. (Not that I've done in-depth testing; after the tiny difference of that cable test I buy large gauge cables and simply don't worry and don't care).

 

[Mad_Economist]

Were these finally posted somewhere?

The full analog recording of the tracks during the video'd ABX is now up! Check the drive for "DAC ABX Live Audio.wav"

We should also see the DAC measurements with these filters used, at the same sample rate as the music files.

Not just freq response but his usual set of DAC measurements - 1kHz, multitone, IMD etc.

For each of the filters used, at the same sample rate of the music files used in the test.

For completeness.

This one may take some time, but I've made Cameron aware of the ask - it's a bit of a time sink to completely match his normal test set, since he has to go through HQPlayer or similar for the upsampling thing, which doesn't to my knowledge entirely work with his existing scripted test routine.

Asking more broadly, would this be more or less of interest to folks than an ABX of two DACs using the mechanical comparator (so no recordings, actually comparing DAC outputs) with in situ recording/monitoring? They might be competing for the same finite amount of time.

 

[SIY]

Asking more broadly, would this be more or less of interest to folks than an ABX of two DACs using the mechanical comparator (so no recordings, actually comparing DAC outputs) with in situ recording/monitoring?

I'd rather see better controls.

 

[Mad_Economist]

I'd rather see better controls.

Let's talk about details then: without me being there physically, what's your highest priority there?

 

[SIY]

Let's talk about details then: without me being there physically, what's your highest priority there?

Best would be to have someone familiar with mentalism or general magic present. Second best would be multiple camera shots set up to show there's no visual cues.

 

[Music1969]

it's a bit of a time sink to completely match his normal test set, since he has to go through HQPlayer or similar for the upsampling thing, which doesn't to my knowledge entirely work with his existing scripted test routine.

Just his normal DAC measurements through HQPlayer measuring DAC outputs.

Just 1kHz tone, multitone, IMD, frequency response - all out of the DAC output and comparing the 2 different filters he used in the test.

He's done this before - its not that time consuming routing through HQPlayer.

The scripted test routine is for playing the music files but I'm not talking about measuring DAC output with music files.

 

[Mad_Economist]

Best would be to have someone familiar with mentalism or general magic present. Second best would be multiple camera shots set up to show there's no visual cues.

I'm slightly confused here - this being a software ABX, where do you want in camera to avoid a potential visual cue cheat? We can definitely get multiple viewpoints with recording on, I just need to know which ones you need to see.

Would him wearing a gopro work well here, as an addition? With that, you should be able to see everything in his POV, in addition to the scope of the external camera(s).

Just his normal DAC measurements through HQPlayer measuring DAC outputs.

Just 1kHz tone, multitone, IMD, frequency response - all out of the DAC output and comparing the 2 different filters he used in the test.

He's done this before - its not that time consuming routing through HQPlayer.

The scripted test routine is for playing the music files but I'm not talking about measuring DAC output with music files.

What levels are you looking for here for the distortion measurements? Just 1khz, an AP multitone, and SMPTE or CCIF would be trivial for sure, but I'm not 100% sure I get what you're looking for. FR is a no-brainer, although you do, of course, get that already from Deltawave - we'll be happy to do that.