Understanding the Role of DACs in a Simple Audio Setup

[old tech]

@gregorio, et al.,
The other day somebody mentioned that you sh/could write an article addressing some (or all) of the common audiophile myths. This current interaction with CopperFox is the perfect example of why I think that'd be a good idea. Yes, I'm sure it would take a lot of your time and probably a fair amount of effort to create, but in the long run I think it'd save you a lot of time and headache. Rather than engaging and arguing with every person battling against the facts, you could simply point them to the article and report them as trolling (or whatever) as necessary.
I think an article like that could potentially be better than refuting everything a person says, because it could be presented as straightforward facts and information rather than feeling (or becoming) personal. Hopefully in the long run, because people aren't feeling attacked and belittled, they'd be more receptive to learning the facts.

Anyway, those are just my thoughts.

Edit: a few further thoughts...
The article could be locked so that it doesn't become a dumping ground for trolling and misinformation.
You (or whoever writes it) could write it in advance before posting it. This way you can take your time without worrying someone is going to start arguing before you're finished.
It could/should include citations.
Each myth/chapter/section could be a new post that could be easily linked later.

The problem is that it is only a small minority that are actually willing to accept that they may be misguided and willing to learn. Most of the fly ins have irrational deep seated beliefs in which they perceive any facts as akin to a challenge to their manhood. That copper dude is a case in point.

 

[megabigeye]

This Fox guy wouldn’t read it. Gregorio writes an article every time he replies.

The problem is that it is only a small minority that are actually willing to accept that they may be misguided and willing to learn. Most of the fly ins have irrational deep seated beliefs in which they perceive any facts as akin to a challenge to their manhood. That copper dude is a case in point.

Yes, but at least he wouldn't have to waste his time and energy replying directly to people who at best don't really care about what he has to say. It'd all be written already. And, importantly for the forum, he wouldn't be feeding the trolls.

I think part of the problem in Sound Science is that people feel personally attacked and belittled and so dig in and become totally dogmatic and inflexible in their ideas and beliefs. If they don't feel like they have to go on the defensive then maybe— maybe!— they'd be a little more open minded.

Anyway, I realize this is completely off topic, so I apologize to the OP.

 

[eq1849]

I don't think you understand what a null test is.

It is not a listening test but a measured comparison of the output of two devices essentially subtracting one from the other so that if they are identical there will be no sound left to hear.

I didn't watch this but I imagine it is at least basically explanatory of what a null test is:

Earlier, I was referring to a null hypothesis of a blind listening test… I was not referring to the Null Test, called the Inversion Test by some.

What exactly do you think the Inversion Test shows?

 

[bigshot]

I’m afraid I don’t know what your purpose is for being here. You don’t have anything to offer and you aren’t interested in anything we have to say. I guess you’re just talking for the sake of talking.

 

[bigshot]

Yes, but at least he wouldn't have to waste his time and energy replying directly to people who at best don't really care about what he has to say.

I think part of the problem in Sound Science is that people feel personally attacked and belittled and so dig in and become totally dogmatic and inflexible in their ideas and beliefs.

I think Gregorio enjoys it. He wouldn’t keep participating here if he didn’t.

And I don’t think anything we do can change the trolls. They’re here to troll.

 

[gregorio]

The other day somebody mentioned that you sh/could write an article addressing some (or all) of the common audiophile myths.

That wouldn’t be an article, it would be a book, a big one. The problem is that there are so many of these audiophile myths and marketers are constantly trying to create new ones, so almost as soon as I wrote the book, it would be out of date. And, not only are new myths created but old ones are recycled from a different angle, requiring a refutation from a different angle.

I did write an article on the 24bit myth (24bit vs 16bit the myth exploded!) many years ago. I got quite a lot of feedback that many found it helpful, it was even cited outside head-fi by a noted expert and it has been useful to just link to it several times but it didn’t really help with those with an agenda, who couldn’t get past their (marketing driven) misunderstanding of the subject or of course with the trolls or shills. There were more than 450 pages of responses and quite a few of them nothing more than fighting the facts.

G

 

[gregorio]

Everyone hears different.

More or less, everyone does not “hears different”. If I play say a Mozart concerto, most people with unimpaired hearing will hear the same thing, some aren’t going to hear it as piece of death metal, EDM or some other sort of sound that isn’t even music. What people perceive, prefer and judge can be significantly different though.

When taking measurements, you're not measuring how your ears are hearing it, so at best any readings are a guide and nothing more.

When we’re measuring say a DAC, then obviously we’re not measuring “how your ears are hearing it” because DACs don’t have any ears and we’re not measuring someone listening to the DAC, we’re measuring the DAC itself. The measurements are therefore obviously NOT a guide to “how your ears are hearing it”, they’re an objectively accurate quantifier of the performance of the DAC itself.

They are a guide to what sounds better. They are not the be all end all of how it sounds.

Again, measurements are NOT a guide to “what sounds better”, they’re an objective measure of the performance of what’s being measured, while “what sounds better” is a personal subjective opinion/preference that exists in your brain and we’re obviously not measuring your brain!

The same is true of all objective measurements, for example when we measure the 0-100kph time of a car; we obviously measure the actual performance of the car itself, we don’t stick electrodes on the head of the driver and try to measure how fast they feel it to be, because obviously we’d then be measuring the driver and not the car and the “measurement” would be meaningless because different drivers are different. If the 0-100kph time was say 6 seconds, then a “measurement” of my mother as the driver would show that the car is ungodly fast, while a “measurement” of a racing driver driving exactly the same car would show it as very slow and most other people somewhere in between. So, what use would such a measurement be? We would learn that the car is extremely fast, extremely slow or somewhere in between, which covers every car ever made, from a Model T Ford to a modern F1 car and therefore it tells us nothing at all! However, if we objectively measure the car itself and it’s objective 0-100kph performance then that is useful, we can accurately compare cars and if we want, decide how fast we personally might feel it to be. And of course, that 0-100kph time would be “the be all end all of how” fast a car can accelerate from 0-100kph, just as a measurement of some aspect of the performance of a DAC would be “the be all end all of how” well that DAC performs that aspect of it’s job.

How is all this not painfully obvious?

Failing(nulling)one of your listening tests, since they are negatively biased, with an expected outcome of null, provides rather weak statistical evidence I think.

Are you suggesting that nulling is strong conclusive evidence?

As you yourself pointed out later, you are confusing the “null hypothesis” with a Null Test, two totally different things and a Null Test is absolutely “strong conclusive evidence”, in fact pretty much the strongest and most conclusive evidence! Incidentally, the expected outcome of a blind listening test is absolutely not the “null hypothesis”, it’s the exact opposite! In science, the whole point of running a DBT and therefore the design and execution of a DBT is to disprove the “null hypothesis”!

The person here who does not know what said test does, and particularly what it does not do is you.

Again, nothing except an unsubstantiated assertion that you just made-up and that is false. A Null Test is very simple, it is effectively just “X - X = 0”, therefore “what said test does” is show when two signals are identical and conversely, when they are not identical and by how much, EG. (X + 1) - X = 1. It really doesn’t get more basic or simpler scientific fact than that, in fact we routinely teach such basic algebra to quite young school children. Unfortunately though, you’ve already demonstrated that a very simple scientific concept is to you “massively complex”!

Fortunately the ignore function is now engaged.

How is it “fortunate” that you’re ignoring basic science in a science discussion forum? The rational, polite and acceptable thing to do, if you want to ignore the facts/science, is not to come to a science discussion forum in the first place, not post BS and then ignore the simple basic science that refutes it!

Nah there are lots of people in the community who are ablle to hold a rational conversation in a coherent manner.

There are indeed but you’re not one of them, or you are able but are extremely rudely choosing not to! Either you’re being a hypocrite or you truly believe that a “holding a rational conversation in a coherent manner” means never addressing the questions/points put to you and instead responding with made-up BS and ad hominems.

G

 

[bigshot]

People seem to not understand that when we talk about fidelity and accuracy, we aren’t talking about taste and preference. Fidelity isn’t a final destination, it’s a baseline starting point. You begin there as your raw calibration, then apply coloration to taste. If you find you don’t like that coloration any more, you fall back to your calibrated point of accuracy and start coloring again.

Without a baseline coloration can wander aimlessly. Random changes yield random results.

If every component was colored differently, every time you replaced one, you would have to recalibrate all over again. And if different sources sounded different, you’d need a different calibration for every source. Chaos!

 

[gregorio]

I was not referring to the Null Test, called the Inversion Test by some.

Yes you were! Your post (#155) was a response to my quoted post which specifically stated “null test” 3 times. Seemingly, you didn’t realise you were referring to a “null test” and instead confused it with a “null hypothesis”. And incidentally, I’ve never heard of a “null test” referred to as an “inversion test” and in audio, a null test is quite a common test.

What exactly do you think the Inversion Test shows?

I explained the simple algebraic basis of what a null test shows in my previous post. Namely, if two signals (say the input and output signals of a device) are absolutely identical or a “difference file” that is not zero (“null”) if they are not identical. IE. The “difference file” is a combination of ALL the differences between the two signals. Because again, X - X = 0, so if ANYTHING or multiple “anything’s” are added or taken away from one of the “Xs” (signals) then that anything/s must be the null test’s output result (difference file).

G

 

[eq1849]

Failing(nulling)one of your listening tests, since they are negatively biased, with an expected outcome of null, provides rather weak statistical evidence I think.

Are you suggesting that nulling is strong conclusive evidence?

I don't think you understand what a null test is.

Earlier, I was referring to a null hypothesis of a blind listening test… I was not referring to the Null Test, called the Inversion Test by some.

As you yourself pointed out later, you are confusing the “null hypothesis” with a Null Test, two totally different things and a Null Test is absolutely “strong conclusive evidence”, in fact pretty much the strongest and most conclusive evidence!

you are doubling down on someone’s(BS5711) incorrect claim and now you are heading into Strawman territory I’m afraid.

 

[2leftears]

They are a guide to what sounds better. They are not the be all end all of how it sounds.
Specially when you're at the TOTL gear and the difference is negligible. One measurement is not inherently better

I think we are talking at cross-purposes; from your response I can tell that I misunderstood what you meant and that you misunderstood what I meant

Of course if measurements show sound reproduction to be different then the measurements themselves aren't the be all end all inference as to which sound is better as far as perception is concerned.

My point what that if measurements show there to be no difference in sound, then any perceived difference is proven to be precisely that; a difference in perception.
Hence before a conclusion can be drawn that two people perceive the same sound different we need to be sure that they are in fact indeed listening to the same sound, which can only be ascertained through measurements (or precise engineering) of that sound. Otherwise the conclusion that people perceive the same sound different could be based on the fallacious assumption that there were actually listening to the same sound whereas in fact they may not have been.

That was the context in which I meant that sound measurements aren't a guide but rather a point of reference.

 

[castleofargh]

I don't see the point of a null test in relation to audibility. The only 2 times that will provide a sure answer is when there is no difference at all(that's valid but impossible to get unless we're checking digital code). And if the difference is so small that the amount falls below anything a human can ever perceive. That second case for me is not that useful as for most audio variables, the individual hearing threshold of that variable will tend to be much higher(as in, we'll stop hearing it sooner) than the ultimate hearing threshold for anything, and the hearing threshold while within a complex signal like music will be even higher than that. So relying on the null test will more often leave us wondering if something could really be audible or not. Instead, a more precise identification of the variables being different through other measurements, would more often give us a more realistic notion about their respective hearing threshold.
So IMO, for everything that does show some difference in a null test and most things will, the null test can rapidly stop being useful.
In the case of DACs, there of is enough of a difference from the clocks so that one signal will start lagging behind the other over time. It's way too small for someone to hear that while using either one of the DACs, and it will be irrelevant for all but the most sensitive time measurements, but by being out of phase, the longer you play, the bigger the signals coming out of a null test. It's the type of false positive we do not want in our seemingly rigorous experiments.


I agree that measurements aren't telling us everything about what we're hearing. Not sure that validates or invalidate anything at all, given how vague it is. But it's true.

I also agree that when defining a signal using n variables, many times we have no sense of equivalence between all those variables(objectively or subjectively). When one DAC does a little better than the second DAC in one measured variable but slightly worse in another, picking the objectively "better" DAC often is a mystery to me.
And while I'm fine with arguments about listening, in practice most people purchase not after listening but after having been convinced by something. That something can be measurements, it can be marketing, or it can be reviews that most of the time just regurgitate marketing. And let's look at the trending marketing approaches those days.
So many are fully focused on 1 variable! Well, that's strange. I'll see someone argue that a frequency response graph isn't telling the all story, also rush to purchase some product which is entirely marketed as having more cowbell(replace cowbell by whatever single variable some marketing is pushing for).
I don't get it. The guy looking for fidelity should probably know better because design engineering is hard enough without having some dude pushing you to only prioritize one variable. I can't begin to imagine how many compromises are made for 1 nonsense obsession that probably was already audibly transparent 1 magnitude ago.
And the guy who cares about hearing shouldn't trust a one-variable pony for the same reason he won't just rely on a FR graph or a THD number.

When we find the weakest variable and focus on improving it, now we're talking intelligent approach. But how often do we see that in the audiophile world? How often can an audiophile think that way, and admit for example that DACs are the most stable, consistent, and accurate piece within their playback chain? It's nearly never what they need to "upgrade" or obsess about when they want better or different sound. And yet, here we are, arguing the audibility of DAC differences as if it was oh so very important of a conversation. If all DACS sounded clearly different from all the others while keeping their measurements, the conclusion would still be that the weakest link is basically always somewhere else(except purchasing a weird design, or having it designed by someone who never thought about a PC being plugged into it).

Yes you were! Your post (#155) was a response to my quoted post which specifically stated “null test” 3 times. Seemingly, you didn’t realise you were referring to a “null test” and instead confused it with a “null hypothesis”. And incidentally, I’ve never heard of a “null test” referred to as an “inversion test” and in audio, a null test is quite a common test.

I explained the simple algebraic basis of what a null test shows in my previous post. Namely, if two signals (say the input and output signals of a device) are absolutely identical or a “difference file” that is not zero (“null”) if they are not identical. IE. The “difference file” is a combination of ALL the differences between the two signals. Because again, X - X = 0, so if ANYTHING or multiple “anything’s” are added or taken away from one of the “Xs” (signals) then that anything/s must be the null test’s output result (difference file).

G

I'm with @eq1849 on this one. He's been talking more or less clearly in the thread and that's on him, but AFAIK, it was always related in some way to having more ultrasonic content. That, no matter what others were discussing. He's a man of focus, commitment and sheer ducking will

 

[bigshot]

I’m the one who originally mentioned that a null (inversion) test proves that DACs sound the same.., and it does. That’s what I thought we were talking about when the word null was mentioned.

https://www.head-fi.org/threads/und...mple-audio-setup.974625/page-10#post-18463499

 

[gregorio]

The only 2 times that will provide a sure answer is when there is no difference at all(that's valid but impossible to get unless we're checking digital code). And if the difference is so small that the amount falls below anything a human can ever perceive. That second case for me is not that useful …

We don’t need to be so strict with the second case though, because an audiophile will not have superior hearing than a child or young teenager and they will not have an high quality anechoic chamber. We’re also not talking about an audiophile doing an ABX where they can loop a short quiet section and whack up the volume, we’re talking about standard/home listening conditions at reasonable listening levels and supposedly hearing brightness, warmth or some other colouration descriptor, which a spectral analysis of the difference file could confirm or refute actually exists.

So relying on the null test will more often leave us wondering if something could really be audible or not.

That depends on how much content a difference file contains and where in the spectrum it is. In the case of doing a null test with say a transducer, I’d agree that it’s relatively useless as the difference file will pretty much always contain content that at least could be audible but in the case of cables and many other components it’s often useful.

In the case of DACs, there of is enough of a difference from the clocks so that one signal will start lagging behind the other over time. It's way too small for someone to hear that while using either one of the DACs, and it will be irrelevant for all but the most sensitive time measurements, but by being out of phase, the longer you play, the bigger the signals coming out of a null test. It's the type of false positive we do not want in our seemingly rigorous experiments.

Sure that can be an issue if you don’t account for it. A difference file with a lot of content does not indicate audibility but a small difference file does indicate inaudibility, as @sander99 stated a few posts back. Ultimately, a Null Test is not an audibility test though, it just tests for the difference and if there’s a significant difference within the audible spectrum then a DBT is the only valid method of determining if it’s actually audible.

I'm with @eq1849 on this one.

It seems I was mistaken that one of his responses was to me rather than another member. It still appears to me that some of his assertions are incorrect, taken in isolation, although I maybe missing the context. If that is the case then I owe @eq1849 an apology in this instance.

G

 

[bigshot]

we’re talking about standard/home listening conditions at reasonable listening levels

This ^

The biggest misconception among both audiophiles (and too often objectivists!) is that every difference is important. Audiophiles worry about things like jitter and cables and other hoodoos that clearly don’t add up to a hill of beans when you’re listening to your stereo in your living room. Likewise, objectivists feel the need to cross every t and dot every i by going into detail on anomalies and exceptions to the rule that only exist in theory. We have whole threads here in sound science that spend 20 times more words describing the unlikely exceptions than the ones describing the rule.

All of this leads to confusion about what matters. Most of the audience for this group aren’t scientists, and they aren’t hardcore audiophools. They’re average people looking for practical advice about how to go about getting the best sound they can when they sit down to listen to music.

It’s all too easy to forget that this isn’t strictly an audiophile forum, and it isn’t a pure science forum either. It’s a forum for applying scientific principles to achieve a kick ass home stereo system. Once we realize that’s the goal, it’s obvious which conversations are productive, and which ones are rabbit holes.

I’m glad every time I see a sentence like the one I quoted above. It shows that the conversation has an anchor to practical reality.