Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?

[DavidEdwinAston]

This is the whole point of stereo versus mono reproduction. Right? Music reproduction is a system, including the equipment, the room and the listener. Our perception of performance of that whole “system” is what matters in the end. Perhaps the top twenty DACs as measured by Amirm will sound exactly the same if swapped out in a specific system of other electronics and in the same room to the same listener. What I am suggesting is to run that analysis of DACs as part of a complete system to provide some objective measure of what many listeners, including myself, claim we are hearing.

Analyzing complete system versus single components adds significant complexity and is its own special form of investigation. Perhaps someone here is interested in taking this on.

in

Errm. Perhaps you are interested in taking it on?

 

[knownothing]

Next up, why don't we measure how dacs affect the timbre of tubas?

Awesome.

You're asking to prove a negative here, that is pointless. Nor did anyone say that it "cannot contribute": It can very well be if the device in question is severely broken.

But if devices give the same output down more than -100 dB, no human can hear the difference. We can't test every human obviously, but there is plenty of evidence to support this claim. To put this into perspective: if a 200m (~650ft) skyscraper would be a 0 dBFS signal, 100 dB down, would be 2mm (~0.08 inch) high. That is how small the differences are we are talking about here.

I think I understand. For reference, where along the continuum of all the DACs measured by Amirm would you start labeling “severely broken”? Any? Half? Curious.

 

[Killingbeans]

For reference, where along the continuum of all the DACs measured by Amirm would you start labeling “severely broken”? Any? Half? Curious.

None of them.

They would have to be broken in a way that changed the dispersion pattern of your speakers, or re-arranged your furniture, or changed the ratio of reflection, absorption and diffusion from your walls.

Impossibly broken.

Talking about the "sound stage" of any gear upstream from your speakers literally makes zero sense.

 

[DonR]

None of them.

They would have to be broken in a way that changed the dispersion pattern of your speakers, or re-arranged your furniture, or changed the ratio of reflection, absorption and diffusion from your walls.

Impossibly broken.

Talking about the "sound stage" of any gear upstream from your speakers literally makes zero sense.

If a sufficient amount of crosstalk was introduced, especially in a non-linear way, that might introduce soundstaging effects. It would be a seriously broken design at that point.

 

[JustJones]

Latest PSAudio dac measurements looks like it's broken

 

[antcollinet]

It’s very complicated and there are a lot of potential variables,

Not really - not when it comes to DACS

A DAC has one, and only one job, and that is to accurately convert the digital representation of music from the source into an analog representation of that music. Well measuring DACS do that with inaudible levels of noise and distortion, and with flat frequency response in the audible band. In other words the analogue output is (audibly) a perfect representation of the digitally encoded music.

If two DACS both achieve this (and well measuring DACS do) then the analog signal from both must be identical within audible limits. By definition, they must sound the same.


Or at least, assuming the amp and speakers are the same, will result in identical sound waves reaching the ear of the listener. What the listeners brain does with that sound information, and how it mixes in the environment, expectations of the listener, mood of the listener etc etc to "colour" the perception of that sound has nothing to do with the performance of the DAC. This is why blind and accurately level matched listening is needed if you want to compare dacs subjectively.

 

[SIY]

Errm. Perhaps you are interested in taking it on?

He knows a relatively inexpensive lab which will do any experiment he requests. I'm getting the feeling that he wants serious people to chase his own misunderstandings gratis, though, and of course no one competent is likely to do that. If I'm wrong, cool, for a few thou he can get all the data he wants.

 

[ahofer]

you have not provided empirical proof that the hypothesis that DACs (or any combination of electronics in general comprising a “system”) cannot contribute to how humans perceive soundstage and localization, given speaker type, measuring equipment and room configuration are held constant, which is essentially the point of the initial post and question.

This is switching the burden of proof. There is a lot of evidence that people can't tell the difference between amps, DACs, sample rates,and wires of all sorts when they don't know what is playing (see "Catalogue of Blind Tests"). There is ZERO evidence of that quality that they can. There is also a lot of audiological science suggesting that the differences among well-measuring equipment measure below the sensitivity of human hearing. The properly formed Null Hypothesis, even before all that evidence, should be that there is no audible difference. Those claiming DACs make a difference should produce a blind, level-matched, controlled ABX test demonstrating it. The burden of proof is on those asserting the Alternative Hypothesis (you), not those saying the Null Hypothesis has not been rejected (most of us here at ASR). And it is VERY telling that the makers of kilobuck DACs have never come up with such a test. Not once. Think about it.

Given the balance of the evidence to-date, it is overwhelmingly likely that the changes you hear do not arise from the equipment, but elsewhere - level differences, room differences, and/or changes in your own mood or physiology.

 

[voodooless]

Talking about the "sound stage" of any gear upstream from your speakers literally makes zero sense.

I would have to disagree here. Any gear that introduces audible frequency response errors can affect soundstage. I doubt it’s generally very severe though.

 

[voodooless]

For reference, where along the continuum of all the DACs measured by Amirm would you start labeling “severely broken”? Any? Half? Curious.

That’s hard to say, because an ordered list of a single measurement figure isn’t exactly useful. You’ll see that not all DAC are recommended. Usually because something is wrong in some way, and this does not always translate to SINAD. Generally though, anything that would produce an audibly different result would probably qualify as severely broken in my book. Most notably are the DACs that have a leaky reconstruction filter. They will not only yield severely mangled high frequencies reproduction, but also have a high frequency droop that can very well be audible. Distortion wise, you’ll have to really mess up, anything with better than 80 SINAD will probably be inaudible, as long as the noise is sufficiently low.

 

[Mnyb]

That’s hard to say, because an ordered list of a single measurement figure isn’t exactly useful. You’ll see that not all DAC are recommended. Usually because something is wrong in some way, and this does not always translate to SINAD. Generally though, anything that would produce an audibly different result would probably qualify as severely broken in my book. Most notably are the DACs that have a leaky reconstruction filter. They will not only yield severely mangled high frequencies reproduction, but also have a high frequency droop that can very well be audible. Distortion wise, you’ll have to really mess up, anything with better than 80 SINAD will probably be inaudible, as long as the noise is sufficiently low.

Yes a filterless NOS DAC is another contender, but are they broken enough?

 

[knownothing]

Not really - not when it comes to DACS

A DAC has one, and only one job, and that is to accurately convert the digital representation of music from the source into an analog representation of that music. Well measuring DACS do that with inaudible levels of noise and distortion, and with flat frequency response in the audible band. In other words the analogue output is (audibly) a perfect representation of the digitally encoded music.

If two DACS both achieve this (and well measuring DACS do) then the analog signal from both must be identical within audible limits. By definition, they must sound the same.


Or at least, assuming the amp and speakers are the same, will result in identical sound waves reaching the ear of the listener. What the listeners brain does with that sound information, and how it mixes in the environment, expectations of the listener, mood of the listener etc etc to "colour" the perception of that sound has nothing to do with the performance of the DAC. This is why blind and accurately level matched listening is needed if you want to compare dacs subjectively.

This is what I said was “complicated”.

“empirical proof that the hypothesis that DACs (or any combination of electronics in general comprising a “system”) cannot contribute to how humans perceive soundstage and localization, given speaker type, measuring equipment and room configuration are held constant, which is essentially the point of the initial post and question.”

My suggestion is that the tools developed by the researchers at MIT provide an opportunity to take out the most subjective element in any attempt to measure the possible contribution of any specific piece of electronics to whole system performance with respect to soundstage, i.e. the human listener, and replace them with repeatable measurements. In such an analysis there are still a lot of variables (complexity) - the entire suite of electronics and connections, the speakers and the room. But holding all but one variable constant should provide a framework to evaluate any component’s contribution to perceived soundstage. Or put another way, if a machine is duplicating human hearing but recording the results, is that perception or measurement?

This essentially provides an objective mechanism to test the null hypothesis that DACs or other electronics have no impact on perception of localization as observed in room, and to prove or disprove the golden ears storyline. You could run a double blind test of humans in parallel using the same set up less the MIT measurements and compare the objective measures to subjective listening reports.

Could be fun.

kn

 

[voodooless]

Yes a filterless NOS DAC is another contender, but are they broken enough?

If you run then at 44.1 kHz sampling, most certainly.

If we talk about timing errors in that case every sample basically has massive amounts of it because it pretends that a sample value is valid for the whole sample. Your basically almost always wrong If anything this just illustrates how bad we humans are at this timing stuf.

 

[knownothing]

This is switching the burden of proof. There is a lot of evidence that people can't tell the difference between amps, DACs, sample rates,and wires of all sorts when they don't know what is playing (see "Catalogue of Blind Tests"). There is ZERO evidence of that quality that they can. There is also a lot of audiological science suggesting that the differences among well-measuring equipment measure below the sensitivity of human hearing. The properly formed Null Hypothesis, even before all that evidence, should be that there is no audible difference. Those claiming DACs make a difference should produce a blind, level-matched, controlled ABX test demonstrating it. The burden of proof is on those asserting the Alternative Hypothesis (you), not those saying the Null Hypothesis has not been rejected (most of us here at ASR). And it is VERY telling that the makers of kilobuck DACs have never come up with such a test. Not once. Think about it.

Given the balance of the evidence to-date, it is overwhelmingly likely that the changes you hear do not arise from the equipment, but elsewhere - level differences, room differences, and/or changes in your own mood or physiology.

We can agree to disagree. But it is ultimately just semantics.

Null hypothesis - Wikipedia

en.wikipedia.org

If you prefer to the construct I am proposing, the Null Hypothesis could be structured such that the MIT system can identify differences in sound localization due to changing one element in the electronics reproducing the sound.

kn

 

[knownothing]

That’s hard to say, because an ordered list of a single measurement figure isn’t exactly useful. You’ll see that not all DAC are recommended. Usually because something is wrong in some way, and this does not always translate to SINAD. Generally though, anything that would produce an audibly different result would probably qualify as severely broken in my book. Most notably are the DACs that have a leaky reconstruction filter. They will not only yield severely mangled high frequencies reproduction, but also have a high frequency droop that can very well be audible. Distortion wise, you’ll have to really mess up, anything with better than 80 SINAD will probably be inaudible, as long as the noise is sufficiently low.

Thanks.

 

[antcollinet]

This is what I said was “complicated”.

“empirical proof that the hypothesis that DACs (or any combination of electronics in general comprising a “system”) cannot contribute to how humans perceive soundstage and localization, given speaker type, measuring equipment and room configuration are held constant, which is essentially the point of the initial post and question.”

My suggestion is that the tools developed by the researchers at MIT provide an opportunity to take out the most subjective element in any attempt to measure the possible contribution of any specific piece of electronics to whole system performance with respect to soundstage, i.e. the human listener, and replace them with repeatable measurements. In such an analysis there are still a lot of variables (complexity) - the entire suite of electronics and connections, the speakers and the room. But holding all but one variable constant should provide a framework to evaluate any component’s contribution to perceived soundstage. Or put another way, if a machine is duplicating human hearing but recording the results, is that perception or measurement?

This essentially provides an objective mechanism to test the null hypothesis that DACs or other electronics have no impact on perception of localization as observed in room, and to prove or disprove the golden ears storyline. You could run a double blind test of humans in parallel using the same set up less the MIT measurements and compare the objective measures to subjective listening reports.

Could be fun.

kn

There are two things you can do - easily on your own, or with little support from a friend (needed for the blind test)

1 - Do a properly controlled blind test - properly level matched - statistically valid, same filter type on each DAC. Prove you can hear a difference. You say you can - should be easy.

2 - Record the output of two dacs. Compare them using @pkane 's Deltawave (it's free to use). Can you hear the difference file? If you can, can you hear it when the track is also playing along side it? You can also use the two recorded files to do an ABX test.


Come back when you've demonstrated you can hear something using one of these methods and we might have something to talk about. Be prepared though to be challenged on your method. You'll need to do it right.

 

[krabapple]

You can bet that tech companies that are creating artificial devices to decipher localization of sounds in the environment understand all the parameters that go into biological hearing systems and are using those as models rather than oscilloscopes or sound level meters. Just saying.

Deep neural network models of sound localization reveal how perception is adapted to real-world environments - Nature Human Behaviour

Francl and McDermott use deep neural networks to reveal the behavioural phenotype of systems optimized for tasks in simulated environments, showing that many characteristics of human sound localization are adapted to real-world environments.

www.nature.com

kn

None of your attempts at sci-ninja support your claim: any item in the chain of gear used to replay a recording that affects the accuracy of timing or level of reproduced sound will affect how we perceive that sound in our listening space.

We can measure timing and level changes to a signal that simply are not perceptible. They will not affect how we perceive that sound.

Qualify your claims.

 

[krabapple]

Are we in Milind Kuncher territory again?

 

[knownothing]

There are two things you can do - easily on your own, or with little support from a friend (needed for the blind test)

1 - Do a properly controlled blind test - properly level matched - statistically valid, same filter type on each DAC. Prove you can hear a difference. You say you can - should be easy.

2 - Record the output of two dacs. Compare them using @pkane 's Deltawave (it's free to use). Can you hear the difference file? If you can, can you hear it when the track is also playing along side it? You can also use the two recorded files to do an ABX test.


Come back when you've demonstrated you can hear something using one of these methods and we might have something to talk about. Be prepared though to be challenged on your method. You'll need to do it right.

Thanks, I can try these.

I would love it if what appears to be the current STOA in terms of non-human sound localization equipment and models could be applied to this question.

kn

 

[knownothing]

None of your attempts at sci-ninja support your claim: any item in the chain of gear used to replay a recording that affects the accuracy of timing or level of reproduced sound will affect how we perceive that sound in our listening space.

We can measure timing and level changes to a signal that simply are not perceptible. They will not affect how we perceive that sound.

Qualify your claims.

OK

any item in the chain of gear used to replay a recording that affects the accuracy of timing or level of reproduced sound at levels within the range and capacity of human hearing will affect how we perceive that sound in our listening space.