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

[zajogungster]

If they don't come in, how are they supposed to be "converted"?
I think some are here because they are in doubt about something. They may agree or be convinced about some subjects but not all. Either it takes some more time or it may never happen.
Don't focus on the negative.

no one talks to anyone here, everyone has some experience and the goal is what can and must be improved in order to achieve satisfaction. There are gaps in sound reproduction and there always will be, but it is unprofessional to absolutely ignore someone's experience, even if he is not a designer, a live-in artist, an engineer. Even scientists don't know what's going on and that's why they're constantly solving it and moving forward. only the unfortunate claims that he is stupid who can hear the difference in sound between a DAC, even if it was high-end at the time, and a DAC for 1/10 the price today. It needs to be solved, what it is, whether the measurements are ok. are they accurate and all variables must be considered. if something can't be measured, you have to communicate!

 

[Blumlein 88]

I had one question : why is everyone so ardent about level matching ??

I mean if one says : well you can’t discern a 15.000 dollar DAC from a 500 dollar DAC because the cheap one is playing louder.. then this kinda proves ( if there even is a difference between them in the first place ) that at least the difference - sound-wise - is so small it can be negated by playing the cheap one a bit louder.

Because it throws a huge chunk of unreliability in your comparisons and causes results which are just not true. When combined with bias from stories about how the product is made or appearance or the belief expensive is almost surely better than cheaper then your listening results range form chaotic to plainly wrong. If a $15 DAC is playing louder it might be heard as better (when in fact there is no difference), or it may be because it really otherwise sounds the same, or it might be despite costing $500 someone has made a design which really is less good. Level matching takes all of that out and lets you determine if these sound same or different.

Personally my reasoning is as follows : I know amplifiers can sound different… if amplifiers can sound different then why wouldn’t a DAC be able to sound different ? It has an output stage, at least this part can make it sound different.. measuring doesn’t tell the whole story. As an analogy : a feeling, an idea is not the same as our neurons firing. It’s something more, hearing music is something different than a measurement.

But if one can’t hear a difference between DAC’s in a (blind) test then there is no difference, at least not for the person listening. Concluding that there is no difference at all ( objectively speaking ) is a bridge too far. You can’t know what somebody else hears. For example I hear nothing above 16khz. If a DAC would make a difference there.. I wouldn’t even hear it.. somebody else might.

While one's hearing might not be a stand in for all people, it narrows the field substantially. With other knowledge of how hearing works and its ultimate limits that can be fairly precise. We don't have people that hear to 50 khz that anyone has ever shown. We do have some that hear to 25 khz, and the effectiveness of that shows it only matters at very high sound levels applying only to a small percentage of young listeners. Of course we also have plenty that from age or other exposure have damaged or diminished hearing. So if such people start claiming differences they hear at 100 khz we have no reason to believe them.

snippage.........

For a 500 dollar DAC : 50 for the plastic case, 150 for parts ( omitting inputs compared to more expensive one ) 300 for the manufacturer , nothing for dealer ( sold directly online )

These numbers are just meant to show that the cost of high end gear is found in other things than the signal path.

Okay as far as it goes, but where is the cost rationality for a $70,000 DAC? There are plenty of reasons one might be a few hundred or even a few thousand dollars for something that in pure sound terms is no better than $150 DAC. But there are limits to what makes any sense beyond just conspicious consumption and irrational beliefs.

 

[Thorsten Loesch]

I have several first generation CD players in my collection and I would be prepared to bet nobody here would be able to pick them from any flavour of the month DAC they want to name in a genuine, blind, real-time, level matched comparison. It's a fantastic reality check to pull out the world's first CD player (CDP-101) and A-B it with some ASR approved DAC and hear absolutely zero difference, 41 years on.

If that mouthful means actually "Audio ABX per Clark/Krueger et at (AABX after this)" I would bet money that it is so.

Yet back in the 90's back in olde blighty (England) HaiFai Rags used to do group tests of the latest and greatest mainly japanese CD-Players, integrated Amplifiers and so on, every month, month on month. Usually at least two SeeDee player group tests at different price points and another two sets for Amplifiers and Speakers.

I was part of the listening panel a fair few times while doing my 2nd degree in London. Listeners were paid, not well, but paid. Hell, I even made money queuing for ticket sharks. And yes, there were consistent preferences for specific products both with individual listeners and across the group of listeners (usually ~ 10). And yes, test were completely blind, level matched etc.

My own later tests used a similar format, because these tests naturally confirmed that there are reliable preferences. I also found interesting that some technical "outliers" showed strong preference polarisations, an example were the Pioneer Legato Link DAC based CD players. They were like Marmite. Many listeners preferred them, but tosome they were the absolute anathema.

The main issue I have with AABX (other than that it cargo cult science [pseudo science], is designed to reliably return "null hypothesis not rejected" in the presence of anything audible but not "day and light" and the fact that is closely related to the fraudulent confidence trick called "shell game") is that no matter the outcome, there is nothing actionable from the outcome for anyone.

If we find that no difference was confirmed (with a 99% likelyhood of the conclusion being in Error, specifically a type 2/B statistical error), what is next step?

If we find that a difference was confirmed (with a 99% likelyhood of the conclusion being correct, avoiding a type 1/A statistical error), what is the next step?

I'm in part an engineer. As such I want something that can be actioned.

An example of something that can be actioned would be: "Out of appx. different chinese 20 dongle DAC's with a mix of DAC Chip's (including Apple, Cirrus Logic, ESS and others) auditioned sighted and scored for preference by around 30 listeners we have a large majority of listeners preferring one DAC Chip, regardless of the brand and looks of the dongles, while other create a mostly negative reaction among listeners" (I had a chance to run this one, it informed the choice of the DAC chip for commercial products).

Another might be:
"Such and such distortion is reliably inaudible and after reaching this threshold concentrate on other parameters, including efficiency, price, profit margins etc.". Similar things are interesting also for other objective parameters. And it would be REALLY useful for consumers if there are clearly established "red lines" where, as long as a product being considered for purchase stays on the other side of the red lines it can be considered "blameless".

In fact, that is normally the regulatory approach to regulating food additives, pollutants etc. et al based on scientific evidence. If we are doing audio scientifically the least sensible expectation would be to have such "red lines" based in reliable science.

So I am not interested in AABX not just because it is likely one of the worst possible tests to confirm or reject subtle audible differences, but because it has no informational value for what interests me in audio.

Thor

 

[Thorsten Loesch]

I had one question : why is everyone so ardent about level matching ??

Very simple. One of the many well known psychoacoustics facts (among many others) is that if two sources have a small but audible level difference, commonly the louder source is preferred by most listeners. There are more like that.

So level differences need to be dialled out, EVEN IN SIGHTED TESTING as one of the most basic measures to remove obvious sources of bias.

Thor

 

[voodooless]

If we find that no difference was confirmed (with a 99% likelyhood of the conclusion being in Error, specifically a type 2/B statistical error), what is next step?

Wait for the next person to take the test. As long as nobody passes, no action is needed. Does this proof anything? No! It just tells you the person taking the test, at that point in time, failed to hear a significant difference. Have a lot of these tests done, and you may deduce some overarching statistical conclusion.

If we find that a difference was confirmed (with a 99% likelyhood of the conclusion being correct, avoiding a type 1/A statistical error), what is the next step?

Figure out how this finding correlates to what can be measured and based on that define a new experiment to prove that new hypothesis. Rinse and repeat...

 

[Talisman]

........

Thor

I started reading your posts with skepticism, then slowly with interest, then with doubt, again with interest..... and then slowly again with doubt, and finally again with skepticism.

You talk a lot but it seems you never get to a point, you keep telling anecdotes, well detailed, well told, sensible, connected to each other. Yet not real proof, never really verifiable important fact.
Now all I see is blah blah blah....

In Italy we call such long and articulated discussions that they do not make any sense "supercazzole"

 

[voodooless]

In Italy we call such long and articulated discussions that they do not make any sense "supercazzole"

I think in English that would be "gobbledygook"

 

[Thorsten Loesch]

Okay as far as it goes, but where is the cost rationality for a $70,000 DAC?

There is non. Where is the cost rationality for a $ 300,000 hand wound mechanical watch that orders of magnitude less precise at keeping time than my $ 200 Bering Smart Traveller which is Bluetooth connected to my phone and automagically adjusts itself to new time zones.

But what you are making is a pure strawman argument, you build up a strawman and knock him down. Well done.

Thor

 

[voodooless]

There is non. Where is the cost rationality for a $ 300,000 hand wound mechanical watch that orders of magnitude less precise at keeping time than my $ 200 Bering Smart Traveller which is Bluetooth connected to my phone and automagically adjusts itself to new time zones.

But the point is that nobody would argue that it does tell the time better. The $70K DAC however is presumed to have the best sound since sliced bread.

 

[Blumlein 88]

There is non. Where is the cost rationality for a $ 300,000 hand wound mechanical watch that orders of magnitude less precise at keeping time than my $ 200 Bering Smart Traveller which is Bluetooth connected to my phone and automagically adjusts itself to new time zones.

But what you are making is a pure strawman argument, you build up a strawman and knock him down. Well done.

Thor

No, not a strawman, not even by your post. $300K watch is not rational either. There are other reasons like conspicious consumption and irrational beliefs. Same reason for uber expensive DACs as uber expensive watches.

 

[Thorsten Loesch]

Wait for the next person to take the test. As long as nobody passes, no action is needed. Does this proof anything? No! It just tells you the person taking the test, at that point in time, failed to hear a significant difference.

No, it does NOT do that. It tells you that person failed the test.

As the "Test" (it is called ABX Test) in "Audio ABX per Clark/Krueger et at (AABX after this)" is basically a confidence trick based on misdirection of biased subjects like the shell game, I cannot accept the results as anything, UNLESS the test and subject are validated by testing an known audible stimulus.

Have a lot of these tests done, and you may deduce some overarching statistical conclusion.

Yup. My overarching statistical conclusion is that the in terms of statistics ABX is extremely heavily weighted towards avoiding "false positives" (statistical Type 1/A error) making "false negatives" (statistical Type 2/B error).

Something else you can do (which I did) is make all individual trials a superset and analyse this.

For example, currently ABX calls for p = .05 and thus needs 9/10 correct. Every test with 8/10 or 7/10 is considered "null hypothesis not rejected".

But what if we have 10 people who all have 7/10 or 8/10 results? That actually is (say) 75/100 and actually easily passes the p = .05 threshold for confidence that a difference was present?

What do we do now?

Wait, I already have your answer.

"Wait for the next person to take the test."

So naturally null results are not rejected in the vast majority of cases. And if we do get a "lucky coin" we invoke statistics again and exclude the lucky coin.

Thor

 

[Thorsten Loesch]

You talk a lot but it seems you never get to a point, you keep telling anecdotes, well detailed, well told, sensible, connected to each other. Yet not real proof, never really verifiable important fact.

I shared my experiences. It is up to you to consider them critically and draw your own conclusions.

It is obviously that you are heavily invested into the "everything sounds the same" viewpoint and you would reject anything I might present.

But let me play along. What would you accept as "real proof"?

Thor

 

[voodooless]

But what if we have 10 people who all have 7/10 or 8/10 results? That actually is (say) 75/100 and actually easily passes the p = .05 threshold for confidence that a difference was present?

You can't just cherry-pick those 10 people and make your new p-value with them. Nor does that new p-value tell you anything about the individuals. If a population has a median income of x, it does not tell you anything about my particular income.

And whatever statistical juggling you do depends rather a lot on the specific question you want to answer.

 

[Mnyb]

no one talks to anyone here, everyone has some experience and the goal is what can and must be improved in order to achieve satisfaction. There are gaps in sound reproduction and there always will be, but it is unprofessional to absolutely ignore someone's experience, even if he is not a designer, a live-in artist, an engineer. Even scientists don't know what's going on and that's why they're constantly solving it and moving forward. only the unfortunate claims that he is stupid who can hear the difference in sound between a DAC, even if it was high-end at the time, and a DAC for 1/10 the price today. It needs to be solved, what it is, whether the measurements are ok. are they accurate and all variables must be considered. if something can't be measured, you have to communicate!

There are many many things in audio that needs further investigation that's very true it's a niche nowadays, so its somewhat under researched currently ,when telephony was invented and then cell phones there was a flurry of research and also around introduction of consumer audio devices and formats during history.
It goes up and down.

But actually sound signature of DAC's is not one of them it's solved since decades. time and resources are better spent somewhere else , return to DAC's when every other avenue off possible improvement are exhausted including having a purpose built perfectly treated listening rooms .There is it's place in the priority list, even more comical it's often the best performing audio equipment most of us own ! and still it should somehow be upgraded to the greatest and latest ? its a very strange idea ? why ?

In a sane world DAC's would be a commodity circuit existing inside other more useful products , like preamps amps and headphone amps and audio processors etc.
And no one would fetishize them as a special product category ? Sadly reality crashes this somewhat as some of the tested combined devices can have bad DAC's . But its usually the most performant part of the product.

 

[SIY]

In a sane world DAC's would be a commodity circuit existing inside other more useful products , like preamps amps and headphone amps and audio processors etc.

In most of the real world, this is the case.

 

[Mnyb]

In most of the real world, this is the case.

Thankyou i forgot where i was today , audio fora is not the real world.

 

[danadam]

But the point is that nobody would argue that it does tell the time better. The $70K DAC however is presumed to have the best sound since sliced bread.

Every time someone brings the mechanical watch example I'm baffled how this difference is not obvious to them.

 

[Thorsten Loesch]

No, not a strawman, not even by your post. $300K watch is not rational either. There are other reasons like conspicious consumption and irrational beliefs. Same reason for uber expensive DACs as uber expensive watches.

Well, there is that.

Now, how about your experience with the watch?

In my Bering watch I change batteries (frequently), the 300k Tourbillion watch you wind by hand. It uses better materials.

Subjectively speaking, the experience with the 300k watch may be to the owner materially different.

As for sound, sound is NOT vibrations in the air.

It is what the individual perceives. And perception is subject to a lot of factors.

But honestly, the whole argument is pointless. Price, looks etc. are not a valid subject of debate if we ask "can device X have a sound signature (that makes it sound different from another similar device and make it less or more preferred to some / many / all listeners and / or myself?"

Thor

 

[yurqqa]

better materials

What are "better materials" ?

Better for mechanical wear? More lightweight ? More legends told about them ?
DO these materials contribute to better acuracy or just subjective feeling of the user ?

 

[Thorsten Loesch]

You can't just cherry-pick those 10 people and make your new p-value with them.

I CAN absolutely do that. As long as I state this in any publication etc..

But I did not suggest that I would cherry pick 10 individuals from a large number of tests (say 100 tests), but that the specific test was taken by 10 individuals, all of which failed to pass the 9/10 statistical threshold but all of which passed the 7/10 threshold.

Thor