CD Transport importance vs. DAC - Page 7

If by "these tests" you mean double blind ABX, then no. It's been pretty well shown what the threshold of audibility is and how it corresponds to, for example, amplifiers. Even a simple Google on "ABX testing" will dig up some articles where clear difference are reliably detected. For example the difference between a single ended triode amp and a solid state amp or the differences between pink/white noise signal with and without a filter added to it. You'll also see that in these same tests that sources with less measurable differences cannot be detected as being audibly different and what the threshold needs to be for something to be audible.

I'm confused. If what he was saying about the tests has merit, and the same tests are used to find these thresholds you are referring to, then how reliable are the thresholds if what he said was true?

This doesn't mean what he said was true, but if you are going to argue from that point of departure it gives you no legs to stand on.

I'm talking about simple tests that the subjects fail, no need to do any double blind testing to make them fail more. And I personally think the whole dbl blind test is invalid myself. Do you cover your ears when you judge HDTV's?
Coke outsells Pepsi, yet in blind tests most coke drinkers preferred pepsi. So which one is better? Well with add'l research it was discovered that drinking a small sample of Pepsi gave it an advantage because it is sweeter. Drinking 12 oz. gave Coke the advantage and people felt more refreshed and many people didn't like 12 oz. of the sweeter Pepsi. Such add'l research in these audio tests is not done because it is more difficult nor would it be recognized as valid.
Thats why I keep harping that long term listening is much more valuable than a simple A-B test. I also stress that long term, the marketplace would discover if these items were all the same and there would only be a few companies making them. That itself far far outweighs any test a very small (percentage-wise miniscule) select group of people may have taken.

If you look at the audio market, whats the thing has the most choices - Speakers, because they are all different even if many are almost identical. However many think you also need to do a blind tests on speaker listening to better determine which sounds better. That is foolish imo because speakers are also furniture that goes in your house and gets looked at everyday. Would you sit in a recliner chair blindfolded to make sure you get the most comfortable one? It would be ridiculous would it not?

The threshold of audibility can be determined (and is) independently of the double blind equipment ABX testing. If you want you can set up the threshold of audibility determination using a double blind ABX, you just have to do it independently of any other testing. For example, take frequency response. You can do double blind testing on a subject with two different frequency sources and determine what the subject can reliably hear. Things like frequency response and harmonic distortion are relatively easy to correlate to the measurements of other equipment under double blind ABX testing.

What I am saying is, that if he is saying that ABX testing is faulty, and ABX is used to determine thresholds, then that doesn't leave us with reliable data to go on, so it is a bad point of departure to argue from with your points. OTOH, if what he says is not true, then there is no point in making an argument whose departure is off of the basis that his argument was true and you have a better argument to work with.

I think you're missing the fact that some people don't fail some tests when they are performed as double blind ABXs. Not only that, the tests where people can pick out the differences using double blind ABX testing correlate well to (among other things) the frequency response and distortion measurements of the equipment under test. We're not trying to prove a null result here, we're trying to determine which equipment specs correlate to audible differences.


Sure it is, the amount of testing done in the areas of audio research is enormous. Your Pepsi/Coke sweeter/less sweet analogy is, for example, similar to 2nd harmonics in amplifiers (sorta). It was not always obvious why people used to prefer tube amplifiers for short term listening tests. Turns out that for any short term test the added 2nd harmonics are almost always preferred by humans. Some people always like them (just like some people always like Pepsi), some people find them fatiguing in long term listening tests.


How so? One thing that has been shown is that most humans have very poor long term auditory memory (not melody memory).


We have discovered that these items are all the same for certain parameters. However, manufactures distinguish their products with other features such as number of configuration options, black vs. chrome adornments, the color of the LEDs, price and immense amounts of advertising designed to convince you there is a difference.

[snip/]


I'm guessing you are intending to claim that many speakers measure identically? If so, that's simply not true; speakers are the one part of the consumer side of the audio reproduction chain that does have the most measurable and audible variation. Combine that with the asthetic compromises that must be made to get truly flat frequency response, low distortion and wide dispersion and there's no way to make even a small range of speakers that will meet the needs of any large population.

I happened to grow up a block away form the original PSB manufacturing plant. I can't even being to tell you how important very careful listening tests are to someone like Paul Barton. Instead I'll just point out this article:

Paul Barton at the NRC

You may want to read the third and second last paragraphs in particular.


For some people (such as myself) the reproductive capabilities of the speakers are far more important than the aesthetics. For others (such as my wife) the aesthetics are far more important. It's nice to have options, even if they require compromise.

Tell me, the last time you purchased a mattress did you perhaps try more than one? Did you close your eyes while you were trying them?

Poor long term auditory memory for what? I would be interested in seeing any studies about long term auditory memory that relate to the ability to recognize differences in the sound of something that is more complex than a single tone, such as a musical passage, a human voice, etc.

For example, I think I'm pretty good at recognizing the voices of my immediate family members (including relatively subtle changes in their voices) even when I haven't heard them for a long term. I don't think I have any special skills in this regard, and I suspect most people also have this capability. Are there any studies that square this common ability with studies showing that "most humans have very poor long term auditory memory"?

Double blind ABX doesn't have to be used, I picked it because if I didn't someone would complain that the methodologies didn't match Double blind ABX testing results correlate well to other methods.

As a weird example, the paper I pointed to earlier in this thread shows that fMRI results correlate to single blind testing for missing fundamental detection (though the fMRI sample sizes were very small). That's sorta neat if you think about it; the areas of the brain that get activated when we hear something or don't hear something actually correlate to the results subjects report in auditory testing.

In any case, my point was (which somehow seems to keep getting missed), that if someone says that the methodology is faulty and you are using outcomes of the faulty methodology as examples of why the methodology works, the argument is faulty.

If the original argument was false to begin with anyway, and ABX testing isn't faulty, then that is the point to argue, instead of trying to prove that ABX isn't faulty by using proof from ABX experiments themselves (you can insert any of the experimental methodologies in for ABX, as the poster did say "these types of experiments" or something to that effect). Otherwise it is a circular argument that can go nowhere...

Things like frequency response and harmonic distortion.


Yeah, I phrased that poorly didn't I? But sorta, yes, at least for the areas that matter to people who claim amplifiers or DACS sound different. In particular, there are tests using "known familiar sources (eg. a family member speaking a familiar phrase)" and distorting it to the point that the subjects can no longer pick the familiar source over a random source speaking the same phrase with the same levels of distortion. With repeated testing the subjects can be trained to continue to pick out the familiar source with higher and higher levels of distortion. Much (depending on the test, all) of this ability is lost over time. If you bring the same person back at a later time and repeat the tests (without the training) the subject can no longer reliably pick out the familiar source. I've also read some reports doing similar training on relative phase, reportedly some subjects are better at detecting relative phase for certain frequency ranges and can be trained to be fairly good at it for short term listening but the skill is not retained well for long term testing. The one thing that would seem to hold out hope for the subjectivist believing in long term testing is that I have read of evidence that training periods get shorter for some of these kinds of testing the more often they are performed.

I'm sorta lost here as to what the original argument was...

My point (which somehow keeps getting missed) is that these (equipment vs. auditory resolution tests) are independent tests. The methodologies don't have to be related, or they can be, doesn't matter, the results still correlate.

I suppose someone could claim hearing tests don't show what people actually hear, but that's why I point at the fMRI results. If you're brain is actually shown to respond or not respond to what people normally claim to hear or not hear how could one continue to claim that hearing tests don't work? You can actually see part of the brain get activated when a signal within normal auditory range is played over the headphones in these tests. Nothing happens when a signal outside of those ranges is played over the headphones.

Then again I suppose you could claim brain activation doesn't matter, but at that point I'm going to point you back to the church of the Great Collapsed Pumpkin Singularity for continued guidance in your quest for audio nirvana. You're no longer dealing in issues that can effectively be discussed on an Internet forum.

We did one. We had an opportunity to see if speakers break in. The dealer in our group was setting up a new pair of B&W 801 speakers in his store. We had another member with a pair that was nearly a year old so we could A/B them at the dealer's store. You guessed it. There was no audible difference between them. Even though the owner of the old pair swore up and down they had mellowed with age, in fact they had not. He paid out $20 in bet money because he claimed he would be able to pick out his own speakers 100% of the time. He managed it about 1/2 the time like the rest of us. So his long term sonic memory wasn't very good was it? B&W's consistency and quality control appeared to be pretty good though.

As far as I could tell, the poster was stating that the testing methodologies being discussed are faulty and don't allow the listener to hear small differences that have been found audible under other circumstances. If you use that methodology for testing other audibility concerns such as thresholds, even if those tests are independent, that makes the results of those tests faulty. So proving those tests aren't faulty by using those tests as examples is faulty logic (as they were claimed to be faulty), instead you need to prove that the tests aren't faulty by other means than the test itself.

The fMRI tests may or may not be a great way to do so for the same reason listed above. I don't know, and I really don't care. I was more interested in pointing out your circular logic so the argument could proceed onward from a more stable platform rooted in something concrete.


It figures you would try to turn your faulty logic into a personal argument with me... You seem like yet another person just looking to argue and always be right at the expense of other people and truth. Oh well, another one for the ignore list.

I guess it goes to show, 20+ years of USENET doesn't necessarily mean someone has learned netiquette.

Don't you think just _maybe_ any comment with a reference to the Great Collapsed Pumpkin Singularity was not designed to be taken particularly seriously?

I was not intending the comments to be personal. If you in fact read them that way you've obviously not had any real experience with netiquette.

Why such extreme sensitivity?

You quoted me, asked me a question. I responded.

One other thing with the testing scenarios that were done, they should also do some testing of known subtle differences and see if that is picked up as well. Something a trained ear would notice but generally be ignored by most.
If you test what you say is not different, then you should also test what is different and get correct results.

Ok, now I'm having a hard time seeing how this isn't a circular argument itself?

Is the proposition that it's somehow possible to hear certain differences in certain circumstances, but at the same time not possible to design a test that would determine if you can hear those differences?


I'm having a really hard time parsing this, your referents are unclear. The are multiple independent (in the mathematical sense) tests (for both thresholds of audibility and component performance); thats the "proof" that that any individual test isn't false. You're never using any of these tests to prove itself (and I don't see any place where I've suggested this.)


Sure doesn't read that way. As far as I can tell you haven't really read what I've posted but instead continue to claim that the arguments are somehow circular. I apologize if I'm missing something but I think you're going to have to explain things more distinctly for me...

I'm sorry, but I'm not really sure how this is really probative of whether a person who is familiar with, for example, a certain musical passage due to repeated exposure can detect a slight change in the sound of the passage, even if it has been a few days since he or she last heard it. The test you're referring to seems to be an entirely different type of test. Indeed, I presume the subjects could determine that the familiar source was being distorted to some degree. The relevant test would almost appear to be the opposite, i.e., what level of distortion was required to reach a point where the subject could state whether the source had been distorted, and more to the point, what impact, if any, did the passage of time have on this ability.

I have yet to see a study referenced that deals with the ability of a person to detect a slight change in a musical passage with which a person is intimately familiar after not being exposed to the passage for a certain duration of time. And I suspect that virtually every test that could be referenced for the proposition that "auditory memory is short" is not really similar to they type of determination we're talking about.

Furthermore, if you play a familiar musical passage on your high end stereo for two weeks, and then you don't listen to that passage at all for the next two weeks, and then you listen to it on your car radio, will the fact that "auditory memory is short" preclude you from determining whether it sounds the same as it did on your stereo? Do you think you would fail that test even if it was blind?

Wow, if that's the type of "evidence" you "scientist/objectivist" types are relying upon, its no wonder there's some disagreements in this thread.

I guess it depends on what you mean by "known subtle differences"? Off the top of my head some of the basic audibility tests I've been involved with or have seen reported include:

- signal levels
- frequency response
- added or removed harmonics
- absolute and relative phase
- noise level sensitivity
- signal levels above or below a given (eg. pink or white) noise floor
- distortion (wave shaping)
- filtering applied to pink or white noise (the ability to detect a notch)
- spatial orientation

Then there are more sophisticated tests for musical memory, beat detection and things having more to do with how we process the basics in combination with each other. (Being a mathematician and a musician I'm always comforted by the results that show a high degree of correlation between these two fields.) Are you referring to the first kind of tests or more to the second kind?


Yup. People do and they do (test and get the expected results).

smitty and dakar: It's not available online, so far as I know, but you should check your local library for the following:

Nousaine, T. "Flying Blind: The Case Against Long-term Listening." Audio, March 1997.

What Tom Nousaine did was to add a load of distortion to a well-known recording, and sent either a clean or a distorted copy to people who owned that CD. He asked them to listen to the copy and decide whether it was distorted or not. People guessed right about half the time. He also gave one subject a standard, short-term ABX test, and got a clear positive result.

ABX tests produce results that are repeatable and consistent with other research. That's why researchers use them. A researcher that used a methodology that was less sensitive would not last long in the field.

Well from a purely musical perspective long term recognition of familiar music when pitch stays consistent rates around 90%. When pitch changes (in the same proportion for all parts of the music) recognition drops to 80%. If the relative pitches of each rise and fall are varied (eg. up and down a semitone respectively) then recognition drops a lot (down to 60% I think) and if both are varied then recognition drops to 50% or so. So what ever tests you want to look at have to take that into account.

Off the top of my head I can't recall seeing any tests results that play with long term recognition of distortion results in music though I'm sure there must be some. Voice recognition is important for various medical and legal reasons which is why I've seen those results.


I've seen both types, I'm sorta glossing over a lot of territory here, I think it would take days and months to cover even a small amount of this topic. One of the perhaps more ironic reasons that the acceptable levels of distortion in long term memory for voice recognition has been tested is our legal system; long term voice recognition is acceptable evidence in almost every court. It turns out that for normal spoken voice, recognition is 95% for familiar voices. People have had to be able to show that long term memory of voices is not altered by various pieces of equipment even when that happens to be a crappy tape recorder, some lossy compression method with lots of distortion, a device with limited frequency response, etc.


Well the ones for pitch and tempo are easy to dig up but I'm pretty sure you want something relating to the types of changes that audio equipment would yield? You may be looking for evidence that non-observable event has not been observed. Personally, I'm pretty confident that the musical recognition abilities would correlate well to the voice recognition capabilities but I can understand if some would think otherwise.


I guess it would really depend on the two pieces of equipment. If you can reliably pick out the differences between the two in a short term double blind ABX you should stand a chance of doing so in a long term ABX. However, everything I've seen reported says that this isn't true. OTOH, I haven't seen a lot of research on this area, I'm sure that's partly because such tests are going to be a real pain to carry out. I'm open to suggestions on how this could be done in some relatively easy but still significant fashion?

you guys are insane if you think all dacs sound the same. thats like saying all ice creams taste the same.

all beer tastes the same too

What do you think could make one DAC sounding different from another?

hmm lets see, the way they are built differently for example? NOS dac? tube dac? oversampling dac? have you ever heard of these?

Tubes are not used to build DACs directly. Rather they are used in the analog sections of devices after the DAC has done it's job. Given the noise and distortion that a tube can introduce devices employing them will often be audibly detectable over devices that do not. This however has nothing to do with the actual DAC portion of the circuitry (though manufactures will often try to obfuscate this fact).

I am quite familiar with the differences between an over sampling and NOS DAC. What part of the differences between those do you think would be audible?

Why yes. Are you also aware that the output levels are also different which can confound comparisons? Normalize those first and then tell me how easy it is.