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

[solderdude]

Hey Frans I actually talked to you on your blog lol. I have a pair of Shure SRH1540 I really like. Do you have a schematic for the cable for these phones?

If it sounds to you like the plot below I can make a filter for the treble. I did not measure one myself but in the treble my measurements are similarish to Sonarworks. The 2 peaks next 2 each other is a challenge though. Reminds me of the HD800 which also peaks at 6 and 10kHz.

 

[ashleydoormat]

I don't really know how it sounds in terms of graph If it's not much trouble for you I'll take any schematic that might work. Thanks!

 

[solderdude]

That's not how proper engineering works.
Take this up via PM as to not go OT.

 

[audimus]

There are powerful skeptical arguments that can introduce doubt into any epistemic system.

Incorrect framing. I have not introduced skepticism for or against the issue of measurability vs audibility or any such lofty thing as nature of knowledge.

Not sure why you are going off on that straw tangent.

I am just clarifying the well known “rules” of science methodology and logic that need to be followed if we want it to be a science-based debate. It is a comment on the form of discourse, not the content of the discourse. It is like telling the bible thumpers who are happy to throw the bible at those that have different beliefs than themselves, that if they understood the bible, what they are doing would be counter to the bible itself.

If that is not done, then like the Bogleheads or the religions, this becomes another religious cult with dogmas that are not critically examined for validity.

Science clearly differentiates what is a conjecture/hypothesis, a theorem and what constitutes a proof and where the burden of proof lies. One of the basic tenets of science is that you do not use a non-falsifiable statement as a conjecture and challenge the other to disprove it and yet this is what is being done in misuse of concepts like null hypothesis here without a proper understanding. I just clarified when such a concept is valid and when it is invalid.

The core message is know what science is before using it like a weapon against a different belief system even if you believe you are in the right.

It has nothing to do with nature of knowledge or the knowability, so I am ignoring the rest of the post which isn’t relevant to my comments.

 

[audimus]

Whilst it is true that loudspeaker and headphone measurements and presentation of results are complex in the case of electronics there is only the electrical connection between boxes and they can only carry an electrical signal all the parameters of which can definitely be measured to a level of accuracy way above audibility.

That is not true as we have already seen. In addition to the nature of the “steady state” responses, there may be boundary conditions that are not explicitly tested for which will have implications on the audibility. The response to a sequence of digital zeros by a device as may happen with content was caught in one review but not in another. That leads to the open question of whether the inputs (pure test tone, a set of tones, etc) we choose for the tests truly represent all the possible conditions we may expect in the usage of that device.

For those who are software developers, this is a daily quandary. Do the test cases used for testing a piece of software sufficiently test every possible condition that is encountered in end usage which will have implication on how that software behaves and judged? The answer is never yes unless it is a trivial program. The devices these days are pretty much like a piece of software.

 

[pkane]

That is not true as we have already seen. In addition to the nature of the “steady state” responses, there may be boundary conditions that are not explicitly tested for which will have implications on the audibility. The response to a sequence of digital zeros by a device as may happen with content was caught in one review but not in another. That leads to the open question of whether the inputs (pure test tone, a set of tones, etc) we choose for the tests truly represent all the possible conditions we may expect in the usage of that device.

For those who are software developers, this is a daily quandary. Do the test cases used for testing a piece of software sufficiently test every possible condition that is encountered in end usage which will have implication on how that software behaves and judged? The answer is never yes unless it is a trivial program. The devices these days are pretty much like a piece of software.

You're overthinking it. Any device can have hidden anomalies that are not easily discoverable. What is being measured and discussed here on ASR is not that. We are not talking about boundary conditions. These tests are designed to reveal obvious, large scale, well-known faults.

It's not enough to keep talking about the possibility of some, yet unknown, audible but unmeasurable effects. Demonstrate one in a well documented, properly controlled test and someone will come up with a way to detect this in an electrical signal. But not before that.

 

[audimus]

You're overthinking it. Any device can have hidden anomalies that are not easily discoverable. What is being measured and discussed here on ASR is not that. We are not talking about boundary conditions. These tests are designed to reveal obvious, large scale, well-known faults.

Not correct. So-called boundary conditions can have steady state audibility effects. If the boundary condition of digital zeros or some threshold used were to occur in the middle of content as silent passages, then the response of the unit would be an audible coloring of the sound depending on what it did with it. This is no different from detecting soft clipping in current tests. So, if you think input for detecting soft clipping is a necessary test procedure, then you must test with inputs that represent forms of silence. This is not academic.

It's not enough to keep talking about the possibility of some, yet unknown, audible but unmeasurable effects. Demonstrate one in a well documented, properly controlled test and someone will come up with a way to detect this in an electrical signal. But not before that.

I have never disagreed with the above.

See the difference between conjectures 1 and 2 in my second post. You are saying that it is 2. The counter to that is conjecture 3, that given any finite set of measurements, there will be a device that will demonstrate something that those measurements will not capture.

All of these are reasonable conjectures but you will never get a resolution between 2 and 3 because they are both non-falsifiable and therefore prone to open ended debate. Unless like Godel, you come up with a proof for the incompleteness theorem and show that any specific finite set of measurements will never capture all possible features of the devices that can be tested in that framework.

This was exactly my point. That the measurement crowd was overpromising that all audible effects were captured in a measurement and the measurement-skeptic crowd was postulating presence of audible factors that may not be captured in any current measurement. The conclusion is that there is no resolution possible to this but one cannot validly pretend that science backs the first and not the second group.

Adding PS: The current set of measurements does test for several boundary conditions in frequency range, input output levels etc and dings units that don’t do well at such boundaries, so it is not that boundary conditions per se are not important but whether we have captured all relevant ones.

 

[NTK]

... That the measurement crowd was overpromising that all audible effects were captured in a measurement and the measurement-skeptic crowd was postulating presence of audible factors that may not be captured in any current measurement. The conclusion is that there is no resolution possible to this but one cannot validly pretend that science backs the first and not the second group. ...

It has been 40+ years since the subjectivist crowd began pushing the idea that there exists audible qualities in audio reproduction that conventional measurement methods fail to detect. This may still be a theoretical possibility. However, AFAIK, no new parameter has emerged as a result of their "assistance" in the research in this area which has passed scientific muster and gained acceptance. I cannot help but to conclude that subjectivists have not helped advancing the scientific understanding of audio reproduction by one iota in the last half century.

 

[pkane]

Not correct. So-called boundary conditions can have steady state audibility effects. If the boundary condition of digital zeros or some threshold used were to occur in the middle of content as silent passages, then the response of the unit would be an audible coloring of the sound depending on what it did with it. This is no different from detecting soft clipping in current tests. So, if you think input for detecting soft clipping is a necessary test procedure, then you must test with inputs that represent forms of silence. This is not academic.

Where did I make a mistake? What I said: "Any device can have hidden anomalies". Do you disagree with this statement? I made no claim about the audibility of such anomalies. Certainly some of these can reveal themselves in a simple, typical test. But then, they wouldn't be considered hidden, would they?

I have never disagreed with the above.

See the difference between conjectures 1 and 2 in my second post. You are saying that it is 2. The counter to that is conjecture 3, that given any finite set of measurements, there will be a device that will demonstrate something that those measurements will not capture.

All of these are reasonable conjectures but you will never get a resolution between 2 and 3 because they are both non-falsifiable and therefore prone to open ended debate. Unless like Godel, you come up with a proof for the incompleteness theorem and show that any specific finite set of measurements will never capture all possible features of the devices that can be tested in that framework.

To demonstrate that two electrical signals differ is not hard or incomplete.

Conjecture "1. Given a finite set of well-defined measurements, there is no audible difference that cannot be captured by that set of measurements" is trivially proven for digital audio. All that's required is that the recording device be able to convert electricity into digital format. That's called ADC, and is the ONLY way audio is captured into the digital format. If ADC can't capture it, it will not be at the output of a DAC in any case, and therefore will not be audible. By definition. Once captured into digital, it is trivial to compare two signals to determine if they are different. I know a little bit about that process

Conjecture "2. For any valid demonstrated audible difference, there is a measurement we can devise to expose that difference" is harder to prove, but follows from Conjecture 1, assuming there is a predictable pattern to this difference. This is the premise of all experimental science. While you can claim that this conjecture is not provable in some philosophical, meta-scientific sense, it is, nevertheless, the basic principle that guides scientific inquiry and progress.

Conjecture "3. For any given finite set of measurements, there can be a device which can be detected audibly but not by that set of measurements." I assume you mean difference? This one implies a limit to scientific inquiry. While there are a few known limits (like Göedel's theorem, or QM uncertainty principle) these are rare in science and require rigorous proof and abundance of evidence when proposed.

 

[Julf]

I am just clarifying the well known “rules” of science methodology and logic that need to be followed if we want it to be a science-based debate.

I suggest you try running through your application/interpretation of the "well known rules of science methodology and logic " again, but instead of audio, apply it to the existence of Bigfoot.

 

[audimus]

It has been 40+ years since the subjectivist crowd began pushing the idea that there exists audible qualities in audio reproduction that conventional measurement methods fail to detect.
...
I cannot help but to conclude that subjectivists have not helped advancing the scientific understanding of audio reproduction by one iota in the last half century.

Cannot speak for the subjectivist crowd since there isn’t some coherent group with an agenda trying to advance something. But considering that for some of these stereotypes, scientific understanding of science is considered irrelevant, they can hardly be held to contribute to it. By the same logic, they could arguably claim that the audio science people have not advanced much of the science to really explain audible phenomena other than show a negative that some differences perceived are not real but that is far from advancing the science to relate measurability to audibility, to explain what measured phenomena correlate with validated perception of depth, openness, detail, etc., instead engaging in “no diffrence” nihilism to hide that lack of progress.

Where did I make a mistake? What I said: "Any device can have hidden anomalies". Do you disagree with this statement? I made no claim about the audibility of such anomalies. Certainly some of these can reveal themselves in a simple, typical test. But then, they wouldn't be considered hidden, would they?

If you read your own statement, you implied that devices have hidden anomalies due to boundary conditions but ASR is not about that, it is about finding obvious, etc.... This is where you are incorrect. Pops from a device are obvious auditory phenomena. The onus is on the measurement to discover and explain such issues. Which goes to the core argument of some that not all auditory perceptions are not necessarily discovered with a finite or current set of measurements. You just bolstered that argument by stating that you would not consider (or ASR would not test for) such hidden anomalies.

Conjecture "1. Given a finite set of well-defined measurements there is no audible difference that cannot be captured by that set of measurements" is trivially proven for digital audio.

This is wrong in that there is no definitive let alone trivial proof. This statement itself is not provable if you understood science because you would have to enumerate all possible audible differences (whether anyone has encountered it or not) and show that the measurement captures it. But it could be a null hypothesis conjecture not a proven theory.

It can be falsified for a specific set of measurements by a single example of an auditory phenomena that is not captured by the measurement.

The behavior of certain digital zeros that result in pops between tracks that was not discovered in a device tested here is existential proof that the above is wrong for this set of measurements.

Now you can say, I will augment that set of measurements once I find that audible difference that is not caught and that will be my next set of measurements. The listener will say, OK define that set of measurement and we will see what audible features are not handled by that. And it goes one but neither will result in a definitive proof. Absence of evidence of an audible difference that cannot be detected in a set of measurement is not evidence of absence of any such difference. So, one can hold on to that unproven conjecture but not claim it has been proved. And the listeners will hold on to their unprovable conjecture that any finite set of measurements will not cover all possible auditory experience.

Your signal comparison is flawed unless you can prove your signal is representative all possible inputs to that device as may happen in real life listening. There is no such proof. There is existential proof to the contrary. The effect of digital zeros in some sequence affecting a device is not captured unless it is accidentally captured as happened in the Emotiva case but the input signals used wee not sufficient for the Paradigm PW link. The problem is not with how much you know the process, it is with the limitations of the methodology and the implicit assumptions you are making which do not show completeness. It is like a software engineer saying, I will show that my code will faithfully compute a function for my inputs but not testing it for all possible inputs that may happen in reality.

Conjecture "2. For any valid demonstrated audible difference, there is a measurement we can devise to expose that difference" is harder to prove, but follows from Conjecture 1,

This as I have pointed out is a valid conjecture but as pointed out above, it is not a provable conjecture since there is no finite enumeration of possible audio differences to show that it is true over audible difference. But it can be falsified by someone coming up with an audible difference that cannot be captured by a measurement. The condition that there is absence of such evidence so far is not evidence of absence of any such difference, so will not prove that unprovable statement. But it is a valid conjecture to hold as long as one does not hold it to be proven true at any time. That is science.

Conjecture "3. For any given finite set of measurements, there can be a device which can be detected audibly but not by that set of measurements." I assume you mean difference? This one implies a limit to scientific inquiry. While there are a few known limits (like Göedel's theorem, or QM uncertainty principle) these are rare in science and require rigorous proof and abundance of evidence when proposed.

I have no idea what you are trying to say here. The last statement makes no sense to me. The conjecture is based on the fact that if there is a device with a audible characteristic that is not manifested in any finite set of measurements, it will be indistinguishable in measurement from another device that is identical to it but without that characteristic and yet they will be different in audibility testing. So to show that conjecture to be true for a finite set of measurement, one would have to find a device who exhibits behavior not captured by that set of measurement.

A trivial hypothetical example. Let us say a simple set of measurements do not detect clipping. A device clips and the effect of clipping is audible. Now take another device that is identical to this one except it does not clip. In that set of hypothetical measurement, both would measure the same but one would be audibly different from the other.

The treatment of digital zeros has already shown this. It does not mean that the conjecture is proved however, because it is unprovable over a non-finite enumeration of possible measurements.

I suggest you try running through your application/interpretation of the "well known rules of science methodology and logic " again, but instead of audio, apply it to the existence of Bigfoot.

I already did to show your misunderstanding of the use of null hypothesis with the unicorn example earlier.

“There is no big foot on earth” is a valid conjecture and a valid null hypothesis because it requires one existential observation to falsify it in a finite space. “There is no big foot in the universe” is a valid conjecture but not a valid null hypothesis the burden of proof now is not on the big foot believer but on you because there is no finite enumeration of all possible universes. I have given the audible difference equivalents of the conjectures to show which are provable and which are not and where the burden of proof is as people seem to be unaware of this distinction even in a science based forum.

 

[ahofer]

I think this discussion misses the mark for most of us. I simply think it is *very unlikely* that there are strictly audible differences between electronics that measure similarly (within, say, a 0.2db range) with current tools- frequency response, IM/harmonic distortion, noise, impedance-matching over frequency, LCR, etc.

I am willing to bet on it. I react skeptically to contrary claims, and indeed many of them appear to be nonsense. I will be grudgingly happy if I am proved wrong, because we will have learned something new-where little new has been learned recently. I will be happy in a more small-minded way if I am not proved wrong.

Certainly the weight of the evidence thus far suggests that measurably similar equipment (as described above) will not be audibly different.

I make no claims to complete proof. I don’t think many others here do either, except perhaps in rhetorical Internet excess.

I dare say most scientific progress has been made along these lines.

 

[GrimSurfer]

I think this discussion misses the mark For most of us. I simply think it is *very unlikely* that there are strictly audible differences between electronics that measure similarly (within, sat, a 0.2db range) with current tools- frequency response, IM/harmonic distortion, noise, impedance-marching over frequency, LCR.

I am willing to bet on it. I react skeptically to contrary claims, and indeed many of them appear to be nonsense. I will be grudgingly happy if I am proved wrong, because we will have learned something new-where little new has been learned recently. I will be happy in a more small-minded way if I am not proved wrong.

Certainly the weight of the evidence thus far suggests that measurably similar equipment (as described above) will not be audibly different.

I make no claims to complete proof. I don’t think many others here do either, except perhaps in rhetorical Internet excess.

I dare say most scientific progress has been made along these lines.

If you were to add "that measure similarly across their common operating ranges", I dare say you'd be right.

 

[pkane]

If you read your own statement, you implied that devices have hidden anomalies due to boundary conditions but ASR is not about that, it is about finding obvious, etc.... This is where you are incorrect. Pops from a device are obvious auditory phenomena. The onus is on the measurement to discover and explain such issues. Which goes to the core argument of some that not all auditory perceptions are not necessarily discovered with a finite or current set of measurements. You just bolstered that argument by stating that you would not consider (or ASR would not test for) such hidden anomalies.

You're wasting time arguing irrelevant points. Nobody ever claimed that ASR tests cover all the possible scenarios. Just the most common ones that can affect the quality of sound reproduction.

there may be boundary conditions that are not explicitly tested for which will have implications on the audibility.

If you want to test for all the boundary conditions, go to town! We'll see you in about... an eternity.

This is wrong in that there is no definitive let alone trivial proof. This statement itself is not provable if you understood science because you would have to enumerate all possible audible differences (whether anyone has encountered it or not) and show that the measurement captures it. But it could be a null hypothesis conjecture not a proven theory. It can be falsified for a specific set of measurements by a single example of an auditory phenomena that is not captured by the measurement.

For digital audio what is captured by an ADC process is the result of electricity being converted into the digital domain. There's nothing to prove here, it's a definition of an instrument designed for this specific purpose.

The behavior of certain digital zeros that result in pops between tracks that was not discovered in a device tested here is existential proof that the above is wrong for this set of measurements.

There is no way to ever know that a device will ever perform its function correctly 100% of the time under all possible conditions. That's true. So, what? Should we stop using DACs, computers, software, cars, planes, TVs, and all other engineered devices? Or give up on testing them for the most common flaws? What conclusion can one draw from your arguments, except for the most obvious and irrelevant one that we can never know everything?

The rest of your post just keeps going over the same point... we can never prove anything to a 100% certainty. This is not worth arguing over.

 

[watchnerd]

Whilst it is true that loudspeaker and headphone measurements and presentation of results are complex in the case of electronics there is only the electrical connection between boxes and they can only carry an electrical signal all the parameters of which can definitely be measured to a level of accuracy way above audibility.
The fact that some people obsess about which chip may be in a box, or what sort of power supply it may have is completely irrelevant, ridiculous even. The "can hear a difference crowd" tend to be unconvinced by double blind tests since they show that they can not.

What appeals to people about different loudspeakers, otoh, is subject to much experiment but so far I have never seen a set of measurements indicating that any speaker has below audible levels of imperfection.

The better tubes I own (for my tube phono stage) have a triode balance match of about 1-1.5%, which would have been considered to be quite excellent back in the day.

And probably better than the tube electronics used to make most recordings in the 1950s, many of which (Living Stereo, Living Presence, Blue Note) are considered to be Golden Age references to this day.

Can I hear that 1% channel balance difference? Hell, no....

The idea that people claim to hear DAC differences below the noise floor on 16bit recording using drivers with +/- 3dB accuracy seems loony.

 

[Thomas savage]

Yes, but you're a master given you started training at the knee of your mum as she strolled the alleys offering quickies for quid...

It's not fair.

I always found it strange how ya dad would insist on maintaining eye contact with me but mum said it kept a roof over our heads and to just be greatful I was not his son..

 

[Julf]

By the same logic, they could arguably claim that the audio science people have not advanced much of the science to really explain audible phenomena other than show a negative that some differences perceived are not real but that is far from advancing the science to relate measurability to audibility, to explain what measured phenomena correlate with validated perception of depth, openness, detail, etc., instead engaging in “no diffrence” nihilism to hide that lack of progress.

"Validated perception" means we can validate that the person believes they have perceived something - even if that something might not exist.

The treatment of digital zeros has already shown this.

What treatment of digital zeros are you referring to?

“There is no big foot on earth” is a valid conjecture and a valid null hypothesis because it requires one existential observation to falsify it in a finite space. “There is no big foot in the universe” is a valid conjecture but not a valid null hypothesis the burden of proof now is not on the big foot believer but on you because there is no finite enumeration of all possible universes. I have given the audible difference equivalents of the conjectures to show which are provable and which are not and where the burden of proof is as people seem to be unaware of this distinction even in a science based forum.

When we comprehensibly measure a piece of gear, and you claim you still can hear a difference, the burden of proof is on you, especially if it is reasonably easy to prove if you can or can not hear a difference.

 

[audimus]

You're wasting time arguing irrelevant points. Nobody ever claimed that ASR tests cover all the possible scenarios. Just the most common ones that can affect the quality of sound reproduction.

This is amusing. On one hand, you are conceding that there are things that are not measured which can have audible differences which is the main point of the audible different crowd. On the other hand, you seem to be implying that the common ones are the only ones that affect the quality of sound reproduction and hence audibility. They are in total contradiction.Pick one. If you pick the former, then you would be right and in agreement with the people who claim there may be audible differences not captured by the current set of agreements. If you pick the latter, then that would be incorrect with the argument we just had for the former.

If you want to test for all the boundary conditions, go to town! We'll see you in about... an eternity.

Why would I want to do that? It is not necessary to show a flaw in the thesis from the measurement crowd. If you want to be practical and say look, we cannot measure everything and there may very well be things not measured that may make an audible difference, you would not be necessarily agreeing that there is such a gap but allowing for that possibility which, from a scientific debate perspective, would be valid and you would be in agreement with people from the other side who are saying the same thing. This is what I pointed out in my very first post.

I am just amused at the apparent dogma people want to cling on to instead claiming science is backing them up while using invalid science methodology (of inference and interpretation].

For digital audio what is captured by an ADC process is the result of electricity being converted into the digital domain. There's nothing to prove here, it's a definition of an instrument designed for this specific purpose.

This is irrelevant to the discussion of what conjectures are provable or not.

There is no way to ever know that a device will ever perform its function correctly 100% of the time under all possible conditions. That's true. So, what? Should we stop using DACs, computers, software, cars, planes, TVs, and all other engineered devices? Or give up on testing them for the most common flaws? What conclusion can one draw from your arguments, except for the most obvious and irrelevant one that we can never know everything?

The answer is very simple. And pointed above. The reason you are in a dilemma and even believe it is in a false dilemma is because you are looking at certain statements as being proven or not. My main point really from the very first post on this is that, science does allow for something called a conjecture. This may very well be true but there is no proof for it. You can even use that as a working hypothesis and proceed on that basis as long as you are open to that conjecture being proved wrong.

The problem happens here (and I don’t mean you personally) is when that conjecture is invalidly considered a proven fact by people who hold it as a dogma and use it to throw stones at opposing conjectures as if science was behind them but not the opposing vconjecture (the equivalent of the bible thumpers who don’t understand the bible) and/or they hold their non-falsifiable conjectures as proven unless the opposing views falsify it. This is counter to what we know as science. That was the only point of my very first post. May want to go back and read it again in that light.

"Validated perception" means we can validate that the person believes they have perceived something - even if that something might not exist.

No, I am only considering in validated perceptions, audible observations that can pass the infallibility tests as designed by the measurement crowd and the audio sciences. Nothing else.

What treatment of digital zeros are you referring to?

The behavior of the Emotiva processor that was caught by measurements and the behavior of the PW Paradigm Link which creates very audible and confirmed pops which can even be recorded (not something one can attribute to some bias or whatever) between tracks which is audible but not caught by this particular set of ASR measurements. I have already explained this several times in the arguments above on why that is important in the context of this discussion and so not going to hash it all over again.

When we comprehensibly measure a piece of gear, and you claim you still can hear a difference, the burden of proof is on you, especially if it is reasonably easy to prove if you can or can not hear a difference.

The above is a strawman not argued against at all. The subtlety of the distinctions made still seems to be eluding you.

There is a difference between a conjecture that holds

1. with any particular finite set of measurements (comprehensive or not is ill-defined and irrelevant), there will be no audible differences heard for two devices that measure the same.
AND
2. this set of measurements capture any and all audible differences

Both are valid conjectures to hold, the first one is a valid null hypothesis where the burden is on the person who claims to here a difference. But the fact that at any point in time, no one has come forward with such a proof is not proof for concluding the second statement above because the first one is falsifiable but not provable. The first one can be falsified by a single example. The audible quality of pops not captured by the measurements on this site with some digital zero inputs for a certain piece of equipment is existential proof that these set of ASR measurements as they exist do not capture all audible qualities.

The second statement is non-falsifiable and while someone may hold this as a conjecture the burden of proof for that will have to be on the person holding the belief. This is the conjecture for which there is a counter and equally valid conjecture that for any finite set of measurements, there will be some audible difference that cannot be measured by that set of measurement. Of course, the burden of proof will be on the holder of such a conjecture just like the counterpart.

By continuing to conflate the two very different conjectures with imprecise statements whether for audio, unicorns or big foot (ignoring snarky intent), you are continuing to miss the point of the very first post on this topic and the subsequent elaborations of the same.

 

[Wombat]

Another long-winded debating enthusiast?

 

[ahofer]

I’d like to go back a bit- how did the Bogleheads get involved in this again? These are people who favor index investing, right? Follow the Vanguard founder John Bogle?