How do you measure sound stage?

[knownothing2]

Did anyone read this article? I don't mean quote mine to cherry pick support for your existing firmly held views. I mean read the article.

I might just eventually read it for my own interest. But if someone else has read it (or plans to), I'd be interested in discussing whether/how it informs the discussion in this thread.

Debates about what you guess it says are valueless for me.

Yes, please read the article. I may be way off here, but it seems this “modeled” approach developed at MIT is moving toward duplicating some elements of how humans perceive localized sound. I thought maybe applying this to the question of whether a single gear change could result in a measurable change in localization of sound in a fixed “listening” position in a stereo/room setting. And doing as @bigshot advocates, and pairing with a well-designed blind AB test with human subjects could add to our understanding and help settle some long standing issues in HiFi. The application of such paired tests could extend well beyond DACs and apply to amplifiers, preamps, phono cartridges and turntables, and even (cringe) cables.

kn

 

[bigshot]

How would a DAC affect localization cues if it’s reproducing with high fidelity? Spacial sound is a function of the room and your brain’s interpretation of physical sound, not signals passing through the gear.

 

[1oad3r]

16/44.1 is designed to be audibly transparent. If a DAC can reproduce 16/44.1 to spec, it’s producing an audibly transparent signal. The only DACs I know of that can’t reproduce 16/44.1 completely transparently are NOS DACs, but those have been obsolete since 1985.

I found this an interesting statement as it doesn’t align with the subjective experiences I’ve had with the many DAC’s I’ve owned. Challenging to locate an authoritative view given the breadth of information available on the Internet, so I thought I’d ask ChatGPT for opinion on a couple of the related questions that have come up on this thread.

In conclusion “….while DACs may handle the same digital file format, their unique architectures and components can lead to variations in sound quality within the realm of human hearing…”

 

[SoundAndMotion]

How would a DAC affect localization cues if it’s reproducing with high fidelity?

It is fun to speculate "how could it", but ultimately your question is number 3 or higher...
1-Can a neural network (meat or silicon) reliably hear a difference in DACs? *
If yes, 2-What are the measurable differences in signals? **
Then maybe, 3-How can those measurable differences be associated with the "perceptual differences? ***

* Assuming, without discussion yet, a proper experimental design.
** Any differences, all differences.
*** An open mind is required to proceed... And "sound stage" need not be the only interesting difference.

 

[knownothing2]

How would a DAC affect localization cues if it’s reproducing with high fidelity? Spacial sound is a function of the room and your brain’s interpretation of physical sound, not signals passing through the gear.

So the original signal has no impact on stereo reproduction? For example in the extreme, the signal sent to the speakers reproducing “Purple Haze” has the same localization cues as the signal reproducing “Green Chimneys” by Thelonious Monk? A ridiculous interpretation of your statement on the importance of the signals passing through the gear, but your comment was general and sweeping.

If the stereo signals in each of two channels are the source of perceived localization of sound, then any element of the reproduction chain upstream that affects the accuracy of the inherent combination of timing, loudness and frequency of the sound coming out of two speakers should affect how a human or machine/model experiences localization cues embedded in the original signal. The underlying differences in execution in two different DACs could occur in the digital circuits, the analog circuits, or both.

I am sure others here are much better qualified to answer your question, but I offer a rudimentary interpretation.

 

[knownothing2]

It is fun to speculate "how could it", but ultimately your question is number 3 or higher...
1-Can a neural network (meat or silicon) reliably hear a difference in DACs? *
If yes, 2-What are the measurable differences in signals? **
Then maybe, 3-How can those measurable differences be associated with the "perceptual differences? ***

* Assuming, without discussion yet, a proper experimental design.
** Any differences, all differences.
*** An open mind is required to proceed... And "sound stage" need not be the only interesting difference.

 

[bigshot]

Localization cues are in the room, not the DAC.

 

[SoundAndMotion]

Localization cues are in the room, not the DAC.

You are out of your depth and simply speak with no authority. First you say it's in the mix and the only relevant measure is channel separation (LOL). You also mentioned the impact of DSP, and now you say the room.
I'd say it's in the vibrations of the 2 tympanic membranes*, the brain's knowledge or expectation of the physical space it's in, and the brain's sensory cues of movement and head position.
*Everything in the signal chain from original recordings to the final transducer (speakers or HPs) can alter the TM's vibrations, so they can all alter the cues, thereby changing what the brain uses to localize individual sound sources. So this does include what you've mentioned... but much more.

EDIT: I'm not saying everything DOES influence localization cues, I'm saying they CAN.

 

[bigshot]

"Everything" can't alter it. The DAC simply processes a digital signal and converts it into an analog one. It doesn't add any depth or directional cues at all. The same goes for an amplifier- it takes a small signal and makes it bigger. None of this involves room reflection, head turning, or brain processing. Regardless of what Chat GPT says, it's simply a matter of fidelity to the original signal, which is easily compared, easily measured and easily determined to be audibly transparent if the distortion, timing and response aren't altered to an audible degree. The body of evidence (again not withstanding Chat GPT) is that amps and DACs are audibly transparent. The differences in specs don't rise to a level that differences can be detected with the human ear.

 

[bigshot]

Clue offered: SECONDARY DEPTH CUES are embedded into the music itself as part of the mix. PRIMARY DEPTH CUES are reflections off the walls and timing changes due to distance caused by room acoustics. The mix is what the sound engineers made it. It can have tons of reverb or be dry- depending on the aesthetic choices of the sound mixer. The room adds a spatial envelope to the physical sound- the same for Thelonious as Jimi. Inbetween the secondary depth cues embedded in the mix and the primary depth cues in the room are the DAC and amp. These add nothing to the spatial signature of the sound. They simply decode a digital signal into analog and amplify it to be sent out to the speakers. Looking there for audible alteration of the spatial characteristics of sound is a waste of time. Current DACs and amps are perfectly capable of passing along a signal unaltered. The mix is also fixed. It doesn't change. It just gets reproduced. So if you are detecting differences in the spacial coloration of the sound, the best place to look for the source of those differences is in the room. It's dead wrong to assume that every link in the chain changes every aspect of the sound. Some of them change sound a lot, and some don't change it at all.

 

[gregorio]

Challenging to locate an authoritative view given the breadth of information available on the Internet, so I thought I’d ask ChatGPT for opinion on a couple of the related questions that have come up on this thread.

Ah but ChatGPT scans for information from the internet. So if it’s a subject that was effectively resolved many years ago but is used in audiophile marketing/reviews, then there will be a lot of mention/discussion on audiophile sites and forums and little or none in professional/engineering/scientific sites or forums. So ChatGPT‘s answer will be largely based on or at least influenced by that and commonly answers with audiophile myths/BS.

1-Can a neural network (meat or silicon) reliably hear a difference in DACs? *
If yes, 2-What are the measurable differences in signals? **
Then maybe, 3-How can those measurable differences be associated with the "perceptual differences? ***

1. Depends on the exact DAC and the test conditions, assuming you’re referring to human hearing. In most cases the answer is “no”. Not sure what you mean by silicon neural network or what relevance that has.
2. The noise/dither, distortion products, anti-imaging filter characteristics and frequency response.
3. That depends on where in the freq spectrum they occur and their magnitude. It also depends on what you mean by “perceptual differences”, we can easily have significant perceptual differences between two signals that are identical (caused by perceptual biases/errors).

So the original signal has no impact on stereo reproduction? For example in the extreme, the signal sent to the speakers reproducing “Purple Haze” has the same localization cues as the signal reproducing “Green Chimneys” by Thelonious Monk? A ridiculous interpretation of your statement on the importance of the signals passing through the gear, but your comment was general and sweeping.

I’m not sure I understand your argument. Of course the difference between two different recordings of two different pieces of music is going to result in relatively massive differences, which are way within the threshold of human hearing. But isn’t the argument about differences in the equipment, IE. The same recording? Different recordings invalidate the comparison. DACs (for example) do not have soundstage, soundstage is not an audio/sound property, it’s a human perception that is influenced by audio/sound properties (such as FR, timing and levels/balance).

If the stereo signals in each of two channels are the source of perceived localization of sound, then any element of the reproduction chain upstream that affects the accuracy of the inherent combination of timing, loudness and frequency of the sound coming out of two speakers should affect how a human or machine/model experiences localization cues embedded in the original signal.

No. If that level of accuracy is below the human hearing threshold then it cannot affect any aspect of human hearing (including localisation), assuming other factors which affect human perception are eliminated. Another important consideration here is your correct assertion of “the sound coming out of the two speakers”. Consider that measurable differences “in the original signal” (digital or analogue) might not “come out of the speakers”, they might be too small to be resolved by speakers (or HPs) or overwhelmed by driver limitations. We can “measure” differences in digital signals to pretty much any arbitrary level, we can measure differences in analogue signals down to the level of thermal noise, or even lower using clever measuring technology (FFT gain for example) but when it comes to sound, we have the physical limits of sound itself, plus the limits of electro-mechanical devices and that’s before we even consider human hearing and the reproduction environment.

The underlying differences in execution in two different DACs could occur in the digital circuits, the analog circuits, or both.

Sure but could those underlying differences in execution result in differences in “the sound coming out of the two speakers” and be of a significant magnitude to be audible?

G

 

[SoundAndMotion]

1. Depends on the exact DAC and the test conditions,

Of course! Note the asterisk.

assuming you’re referring to human hearing. ... Not sure what you mean by silicon neural network or what relevance that has.

No, I'm not restricting the idea to humans. Neural networks (see wiki) are amazingly good at pattern recognition, including using audible cues to detect location of one or more sound sources, They can be biological ("meat") or mathematical ("silicon"), as in the MIT paper cited above. Is the relevance clear?

In most cases the answer is “no”.

Agree as to "most", but those are the less interesting in the context of this thread's discussion, IMHO. Although very unlikely, it would be exciting if 2 DACs that should be audibly indistinguishable, can be distinguished. If the unlikely does occur, understanding how would contribute to the understanding of auditory perception.

2. The noise/dither, distortion products, anti-imaging filter characteristics and frequency response.

Well yes, those are good places to start... but I would also look closely at one of your faves: the residual signal after "nulling".

3. That depends on where in the freq spectrum they occur and their magnitude.

You would certainly look at that, but IME the associations will be more complex, and fall under spectral (as you say) and temporal pattern recognition... fun to speculate, but for me only relevant after questions 1 and 2. Note, also IME, magnitude may defy expectations.

It also depends on what you mean by “perceptual differences”

Reliable ability to distinguish 2 DACs (as one example target among others)

we can easily have significant perceptual differences between two signals that are identical (caused by perceptual biases/errors).

Not with good experimental design. "easily" and "significant" should be minimized or eliminated.

 

[bigshot]

Although very unlikely, it would be exciting if 2 DACs that should be audibly indistinguishable, can be distinguished.

If indistinguishable can be distinguished, it isn’t indistinguishable, is it?

 

[SoundAndMotion]

If indistinguishable can be distinguished, it isn’t indistinguishable, is it?

Very good! Notice I said "should be" not "are". Can you distinguish that?

 

[71 dB]

Regardless of what Chat GPT says,

Didn't Chat GPT go insane for a few hours the other day speaking gibberish in Spanglish?

it's simply a matter of fidelity to the original signal,

If we simplify things then yes, but some kind of violations into fidelity may not be that detrimental for soundstage because stereophony is an illusion to begin with. An altered illusion is still an illusion. For example vinyl with its distortions and elliptic filtering seems to not do much bad for soundstage.