GoldenSounds passes apparently ABX test for DACs (NOT Really)

[oleg87]

GoldenSounds ran his ABX tests with a Holo Audio May DAC—his preferred DAC. I noticed he was quite critical of the Denafrips R2R DACs he reviewed, particularly the “misleading” NOS mode.
Not surprisingly (in a good way!), he appears quite knowledgeable about R2R DAC technology.

Is the May DAC a potential factor here? Can some “glitch” exacerbate or transform the high-freq small differences into something much more audible?

EDIT: And is the 176.4 kHZ sampling a compounding factor for a R2R DAC?

Not unless the May has kind of (very) strange implementation bugs. Again, we're talking about comparing 2 audio files at the same sampling rate with absolutely miniscule differences.

 

[danadam]

It's strange that the video never describes how you can distinguish between the different reconstruction filters in terms of sound.

I see no reason why one could not replicate the test with a video camera capturing the whole event

So, for example, show the test procedure in the video (perform the ABX-Test before the camera) and give a description of how others can also replicate the result, i.e. what to look out for in the music sample and which is the best place in the music sample to differentiate.

It’s a shame he didn’t perform the test in his video like he made out he was going to at the start.

Um... like here?

 

[Ken Tajalli]

Um... like here?

And if they had paid attention to the first video, they would have known of its existence.
Thank you.

 

[ctrl]

Um... like here?

Yes thanks for the link, that is what I have missed in the first video (or not have recognized the hint to the second video). With this information one has a starting point to replicate the ABX-Test with a change to hear the difference too.

I will explicitly refer to the second video in my two posts #241 and #251 where I show how to pass the ABX test while cheating, so that it is clear that Goldensound in the video is not cheating.

 

[Sokel]

Um... like here?

Seems like he did his homework and we didn't

(no wonder,it shouldn't take more than reading a couple of threads here to cover all bases)

 

[amirm]

Seems like he did his homework and we didn't

I watched it when it was posted. That is where I got the information about his filter types and such.

While on this, note that he is able to pass this test due to rapid switching. This dismantles the myth that you need "long term" listening to detect these differences. Pretty sure he would fail this test if he listened to one track. Waited a day and then listened to the other track.

 

[Tell]

We have the files, so we can just see that. "correct" is not an easy definition, because various filters make various tradeoffs. Given that, however, I would consider both of these filters transparent enough. You can call that "correct" if you like. Apparently, for Goldensound, they are not transparent enough, though

You can see them at the Deltawave tests early in the thread.Nothing exotic,rather normal I would think.

Yeah I have seen the spectrums of both files, but I don't think that the upsampled one looks natural. It looks like any other track with a pink noisy shape to it all the way up to 20khz where it goes down hard as it should from the ADC AA filter, but then at 20.5khz the slope slowly starts flattening out until 21khz where it's almost completely flat shelf up to 22khz and then just steeps down. Sure you might be able to hear whatever is going on if you have hearing like a 10 year old like Goldensound apparently has, but what kind of information is there really? Is that something that should naturally be there in the recording? I don't think it is so I'm not sure if I would call that a correct filter because of that wonky shape.

 

[ads_cft222]

I watched it when it was posted. That is where I got the information about his filter types and such.

While on this, note that he is able to pass this test due to rapid switching. This dismantles the myth that you need "long term" listening to detect these differences. Pretty sure he would fail this test if he listened to one track. Waited a day and then listened to the other track.

One could argue that long term listening of this track using this filter mapped the expected sound output in his brain thus making it easier to detect . As an analogy solving the same type of math problem many times makes you quicker as you automatically recognize a pattern thus your brain conserves energy and time without thinking from scratch …

 

[sonitus mirus]

One could argue that long term listening of this track using this filter mapped the expected sound output in his brain thus making it easier to detect . As an analogy solving the same type of math problem many times makes you quicker as you automatically recognize a pattern thus your brain conserves energy and time without thinking from scratch …

Or, it might be unrelated to the audio of either tracks and some anomaly of the software or hardware when switching quickly from one file to the other.

 

[dasdoing]

Hard to believe that even when his ability to hear this high is confirmed he could differentiate two audio tracks because of the roll-off

 

[Sokel]

Or, it might be unrelated to the audio of either tracks and some anomaly of the software or hardware when switching quickly from one file to the other.

Foobar would show that,every glitch of the hardware when is in ASIO driver shuts the playback,it somehow ties it to it.

 

[solderdude]

Maybe its an idea for people with hearing not reaching 20kHz (which will be the majority of audiophiles) to slow the sound down the song to half speed so that 20kHz becomes 10kHz and then compare the files yourself. Then it would be easier to understand if and how much the files differ at normal rate ?

 

[oleg87]

As I expected, I slowed down the speed to 0.25x to place the differences well within frequencies I can hear but still soundly failing the ABX test.
(edit: haha, solderdude beat me to it)

 

[Ken Tajalli]

Maybe its an idea for people with hearing not reaching 20kHz (which will be the majority of audiophiles) to slow the sound down the song to half speed so that 20kHz becomes 10kHz and then compare the files yourself. Then it would be easier to understand if and how much the files differ at normal rate ?

What does that prove or show? that there is a difference in ultrasonic bandwidth of the tracks? we already know there is.
Is that all we have? it must be the 21kHz he can hear!
could it not be something else we have not discovered yet?
The guy possibly can detect 21kHz at extreme levels! A pure single tone!
In music, it is mixed with all sorts, at levels of -40dB or less, he couldn't possibly be hearing 21kHz at -40dB, let's be realistic.
The answer is elsewhere.

 

[solderdude]

The test I proposed is to see if one can hear the uploaded music samples (he used music not test tones) and bring the disputed part of the frequency range into the range where even old farts can detect differences and then ABX to see if one can detect differences with that recording.

It is not to debunk anything. Just a method that could be used.

I'm not fussed myself. I can't hear it anyway (I am DAC-deaf)

 

[Ken Tajalli]

I'm not fussed myself. I can't hear it anyway (I am DAC-deaf)

It's all that solder fumes he have inhaled!
Do you also have thick skin on your thumb and index finger, like me? from repeated burns?

 

[sonitus mirus]

What does that prove or show? that there is a difference in ultrasonic bandwidth of the tracks? we already know there is.
Is that all we have? it must be the 21kHz he can hear!
could it not be something else we have not discovered yet?
The guy possibly can detect 21kHz at extreme levels! A pure single tone!
In music, it is mixed with all sorts, at levels of -40dB or less, he couldn't possibly be hearing 21kHz at -40dB, let's be realistic.
The answer is elsewhere.

I agree, I think. The assertion that someone of his age can discern subtle differences around 21kHz is quite extraordinary. If there is indeed a legitimate tell, it's more plausible that it stems from factors other than the ability to hear differences at such high frequencies. It appears we might be jumping to conclusions based on a very limited set of data, which on its own may not be dependable.

 

[sonitus mirus]

The test I proposed is to see if one can hear the uploaded music samples (he used music not test tones) and bring the disputed part of the frequency range into the range where even old farts can detect differences and then ABX to see if one can detect differences with that recording.

It is not to debunk anything. Just a method that could be used.

I'm not fussed myself. I can't hear it anyway (I am DAC-deaf)

Introducing test results from different sources is a solid way to double-check the initial findings, especially since there's some history suggesting the first tester might have had a bias. By conducting these tests on various kinds of equipment, all using Foobar ABX, we can see if the differences noticed are consistent. This approach not only makes the results more trustworthy but also helps us identify if certain equipment is causing issues that weren't clear before. Expanding the testing like this helps overcome potential biases and clearly shows whether the differences heard are truly about the audio quality or just quirks of how the tests were set up.

 

[amirm]

One hassle here is him picking an oddball DAC to do this test on. We don't know if that DAC generates any kind of "tell." Or an easy way to check his testing.

 

[MacClintock]

Um... like here?

I guess nobody here would pass the ABX test. So either he has really amazing listening capabilities or he is fooling us big time. It is just so hard to believe that any information above 20 kHz should be audible to any human, especially at this low SPL, masked by music. I do know that I will save my money and not get into Chord sharp filter DACs, as it would be lost on me.