GoldenSounds passes apparently ABX test for DACs (NOT Really)

[VientoB]

Out of interest what is the level of the signal Cameron picked up and at what volume level was he listening to the piece of music?

 

[Mad_Economist]

Out of interest what is the level of the signal Cameron picked up and at what volume level was he listening to the piece of music?

That's a very good question - I don't have an answer for it, because Cameron doesn't (yet) have a calibrated ear simulator, so we can't verify his acoustic levels. That's something we'll have to get back to you on roughly in 3 weeks, since that's about when he should have an ear sim with a known sensitivity in dBSPL/V.

 

[RandomEar]

So there are some misconceptions here, I think:

• That's the average of the entire recording, not the peak - the peak value will, of course, depend on the length of the FFTs we use, because you're effectively integrating over the entire duration of the FFT, but something that's inarguable is that an FFT of the entire audio sample does not reflect the peak value.

True. The numbers were more about having a ballpark figure. It would be more precise to specifically look for peaks in HF content >20 kHz and see what else is present at that point in time and compare those SPLs.

• -35dB for a distortion product would be quite audible with a pure tone, so long as it was not an ultra low frequency and/or the harmonic wasn't H2 - it's important to bear in mind that there's no universal threshold for distortion audibility outside of the actual threshold of hearing, because it's entirely dependent on the masking effect, and masking is a frequency and level varying effect

This isn't about pure tones, though. The test sample we're talking about is music. With actual music and a broad distortion spectrum, most people still fail to identify -30 dB or less. Yes, that's not the limit of audibility. Yes, trained listeners perform better. Yes, it doesn't cover all types of music and distortion and doesn't directly apply to this ABX test. It's just a ballpark figure to get an idea how much smaller a signal typically has to be to possibly consider masking.

• Distortion audibility isn't a good analog for audibility of portions of music, since distortion is tautologically a product of another signal, which, when distortion is a negative decibel value, is tautologically a stronger signal. The sections of high frequency information that are captured with the wider bandwidth filter are short, transient sounds which do not have to have a larger masking signal present - to give an example here, I can get a -110dB average value with a sample of a gunshot in a long period of silence, but it's still going to be audible for the moment it occurs if you can hear the frequencies it covers

As stated, distortion was an example to get an idea of the order of magnitude we're talking about. For the sample files we're talking about I just checked how much difference there is between 3 kHz and 21 kHz in a region of ~0.1 s around some of the transient "ping" sounds, which do have the highest HF content. It is >=30 dB of difference. Your point about short peaks getting ironed out by the average is correct. As far as I can tell from the analysis, my guesstimate of SPL difference still happened to be correct in this case. Probably just luck.

And I'd like to reiterate that - as far as I can tell from the given information - Cameron barely manages to identify tones at 21 kHz under ideal conditions (which is impressive enough). I'm just sceptical that he can identify content in this frequency region while listening to actual music with so much other "dirt" around at the same time. I'm not saying it's impossible, I just like to see clear evidence that this is the reason he aces that ABX test. And I'd like to understand other possible factors as well, in case there are any.

 

[Mad_Economist]

True. The numbers were more about having a ballpark figure. It would be more precise to specifically look for peaks in HF content >20 kHz and see what else is present at that point in time and compare those SPLs.

Deltawave has a supporting function natively! It's the "Δ Waveform" tab - looking at the spectra at ex. 8 seconds or so would likely prove interesting.

The overall delta waveform peaks at -46dB, I'll have a look at the spectrum for a short FFT in that time window

I'm just sceptical that he can identify content in this frequency region while listening to actual music with so much other "dirt" around at the same time. I'm not saying it's impossible, I just like to see clear evidence that this is the reason he aces that ABX test. And I'd like to understand other possible factors as well, in case there are any.

I'd agree with this, and hopefully we'll be able to pin down the non-malfeasance potential factors for folks. Obviously, can't really do much about the perception of potential malfeasance, but if there's anything additional you'd like beyond the analog recordings Cameron will be uploading when he's back from Munich, please let me know with a mention here and I'll try to get it added.

Edit: threw it into REW and windowed off the section of the largest peak with a 100ms long window. The highest level content above 10k hits around -87-90dBFS, while the highest super-20.5khz content is in the area of -111-116, so we're looking at -26-32~dB of difference, assuming that 8khz isn't producing significant masking above 20khz (anyone got a masking estimate for 8khz handy?)

 

[Blumlein 88]

I also provided a file which slowed down the file by 50% and filtered out the stuff below 20 khz if anyone wanted to look at it. I am not sure it is still available for download.

 

[Somafunk]

Keep this up please folks involved, it’s an interesting conversation which by page 40 of a thread is a rarity.

 

[Pinkynator]

Is there a (part of the) video where we SEE him do the test?

 

[Mad_Economist]

Is there a (part of the) video where we SEE him do the test?

Yes! The full 20 guess run is linked in the description

 

[krabapple]

I would still call that a 'tell' rather than an appreciable difference. It does prove that the tell can be heard though (could be noise, an artifact or whatever)
It just may not be the thing that makes either tested subjects more enjoyable than the other.
The latter is usually what is claimed by people with 'superior ears/gears' and that the difference is so night and day that they don't even have to A-B it.

An AB(X) can be passed successfully with 'tells' in it a few times with statistical relevance yet can be inconsequential with music for the vast majority of people.

Indeed. This cannot be emphasized enough. It's not relevant to the claims typically on offer.

"I could hear the difference right away"/"it wasn't subtle"/"even my spouse"/"creamier highs"/"deeper bass"/"more resolving"/"wider soundstage"/"better imaging"

!=

"I focused on one cymbal hit at 2'15" and aced the ABX "

 

[krabapple]

To realistically hear a difference, the two samples must differ by at least about 0.5dB over 1-2 octaves, i.e. at least from 10kHz to 20kHz 0.5dB difference.

IIRC , in the most sensitive range, a difference of ~0.2-0.3 dB is detectable. I don't recall if that is with 'test tones' or with music, or speech.

 

[krabapple]

If the test is a true and legitimate result, then it's evidence that some people can detect differences between certain digital reconstruction filters if they turn the volume way up in the highest audible octave - yes?

Did Goldensounds turn the volume way up in the higest octave, or did he play files at 'normal' level but use headphones that were 'flattish' out beyond 20kHz,?

Or did he do both (turn up highest octave + use atypically responsive headphones)?

I'm not getting clarity from the discussion.

 

[krabapple]

Wasn't that AES paper highly criticized?

In the comments section of the article itself, we find, a decade or so ago, Amir himself plus Bob Stuart, arguing with a reader who questions the paper's claims. And Arny Kruger chimes in too!

AES Convention Papers Forum » The Audibility of Typical Digital Audio Filters in a High-Fidelity Playback System

The AES Forum is where our members get to interact with each other in between conferences and section meetings.

secure.aes.org

 

[krabapple]

Where are the DAC output measurements for GoldenSounds test files?

Were these finally posted somewhere?

 

[amirm]

My definition of a "tell" is a cheat that is outside of what is being evaluated. Finding a specific segment in a track where the difference is audible as such, is not tot a "tell." Let's say I have a speaker that distorts with deep sub-bass content. When I play something with that precise spectrum and distortion becomes audible, that is not a cheat or a tell. Indeed, I have such a track that shows problems that almost none of my standard test tracks show. We want to know about these problems and not close our eyes and dismiss them as being a tell/cheat.