Home |
Search |
Today's Posts |
#1
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
The classic abx listening test is a chore to setup and perform. Its
clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. -=Bill Eckle=- Vanity Web Page at: http://www.wmeckle.com |
#2
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"William Eckle" wrote in message
om... The classic abx listening test is a chore to setup and perform. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. And others will claim they can hear it. Perhaps an ABX test under this circumstance with some training might be of some value. But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. |
#3
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
It's certainly not a new method, but it's nice to have new software that
automates the process. ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#4
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 28, 11:21 pm, "Harry Lavo" wrote:
Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, This is also false. where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Where you won't find it. But if anyone wants to read what Greenhill actually found (rather than someone's re-invention of it), e-mail me and I can send you the article. Harry is peddling pure pseudoscience here. bob |
#5
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. .....and then,once these 'differences' have been 'heard' to the listener's satisfaction, it's time to see if they're real...via a blind comparison. You just can't get around that. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Nonsense, Harry. ABX has been used to detect the difference between .wav source and 320 kbps mp3s -- which are *extremely* difficult to tell apart. Sensitivity is certainly reduced for comparisons (sighted or blind) when music is used RATHER THAN TEST TONES. This is not a function of sighted versus blind, and the Greenhill tests do not say otherwise. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. And others will claim they can hear it. Perhaps an ABX test under this circumstance with some training might be of some value. But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. The main value of a difference test is where it produces a null (and by null, it must mean residual levels below the practical or theoretical limits of human hearing). There, subjectivists would have to invoke some new sort of audible effect, akin to homeopathy, where vanishingly small amounts of 'medicine' are said to effect cures. I wouldn't put it past them. ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#6
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Harry Lavo" wrote in message
"William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. There ain't no such thing. The world is full of differences that are measurable but not audible at all. So, one caveat is that the difference has to be audible, and not all differences are audible. Secondly, the audibility of many differences are contingent on a wide variety of influences, two of the stronger ones being the sensitivity of the listener, and some properties of the music being used for the comparison that may be non-obvioius. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Or, you picked the wrong music or associated components. Or, the listener's sensitivity was atypically poor for any number of different reasons. Unfortunately, this is not true. So far so good. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. Ironically, these same listeners would no doubt report a plethora of audible differences in a sighted evaluation. ;-) The test is as much a test of the listener as it is of the items under test. Name a subjective test methodology where this *isn't* true. Some people like to play a silly game where they assert that a certain general problem that afflicts just about any test is a problem of just their least favorite kind of listening evaluation. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. where easily identifiable distortions can be used at various levels of impact to "train" listeners, Surely true for more than just codecs, so this is also a false claim. who then listen to the blind samples but "knowing" what they are listening for. Which opens yet another general problem with subjective tests. Differences are very often easier to detect if you know what to listen for, so how do you know what to listen for until someone has first actually heard it? Open-end evaluation of audio gear does not work this way. Actually, open-end evaluation doesn't work at all. It is open-ended all right - it contains poor-to-non-existent safeguards against false positives. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. A truism that all by itself can't support any particular viewpoint. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Fact is that lots of people report fatigue while listening by *any* means that efficiently rejects false positives. The reason is pretty obvious, most of their former experiences were all about false positives. Take those away and they suddenly have to produce more than fond wishes that they heard a difference. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. There ain't no such rule. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. This is one of those statements that immediatly raises a red flag. Broadband white noise, being an unweighted varying and random combination of all frequencies, is very much immune audible effects due to frequency shifts. Theresfore it is a nonsense statement because it is contingent on something that doesn't ever happen. As soon as it comes to listening to music, sensitivity decreases or disappears. Name a subjective test methodology where this *isn't* true. This isn't just speculation No, its mostly false and mistleading claims, that shouldn't even qualify as reasonable speculation. ....review the Greenhill tests in Stereo Review (search index at their site). Greenhill was an advocate of ABX at the time those tests were written and published. I spoke with him personally for hours in those days. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. Been there, done that many times. The problem with this methodology is that it is easily overwhelmed by trivial differences such as phase shift of any kind but linear phase. You end up with two waves that sound very much alike, but generate relatively large difference signals. The irrelevant information in the difference signal masks the difference that was being listened for. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. This can easily happen, because test equipment has become very, very sensitive. And others will claim they can hear it. You may hear a large difference signal, but is it all that relevant? Perhaps an ABX test under this circumstance with some training might be of some value. You can learn how to do that at www.pcabx.com . But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. www.pcabx.com has links to everything you need to do high quality listening tests as easily as possible, given the inherent problems with listening tests in general, such as the ones I mentioned above. |
#7
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Arny Krueger wrote:
"Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. There ain't no such thing. The world is full of differences that are measurable but not audible at all. So, one caveat is that the difference has to be audible, and not all differences are audible. Secondly, the audibility of many differences are contingent on a wide variety of influences, two of the stronger ones being the sensitivity of the listener, and some properties of the music being used for the comparison that may be non-obvioius. However, it only takes one positive ABX test run to demonstrate that the person who took the test could differentiate the sound of A and B. (Assuming of course that the test was set up properly.) And it only takes one person demonstrably hearing a difference, to 'prove' that the two things sound different. That's not the same as saying 'anyone' will be able to hear the difference, of course. ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#8
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"bob" wrote in message
... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, .This is also false Wishful thinking. where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. More evidence (admittedly anecdotal) than has been presented to validate that it works for open-ended evaluation of audio components. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. To paraphrase, you haven't a shred of evidence that long term, exploratory tests paired with short term comparisons doesn't overcome this limitation. We are talking MUSIC, after all...not white noise. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Where you won't find it. But if anyone wants to read what Greenhill actually found (rather than someone's re-invention of it), e-mail me and I can send you the article. And if you wish to email me I can send you an accurate and complete Excel table of the results. Harry is peddling pure pseudoscience here. You've heard from a true ABX believer. Ask for the validation test that this technique, developed very specifically for codec distortions, works as the best tool for open-ended evaluation of audio components. |
#9
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Steven Sullivan" wrote in message
... Harry Lavo wrote: "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. ....and then,once these 'differences' have been 'heard' to the listener's satisfaction, it's time to see if they're real...via a blind comparison. You just can't get around that. Sure you can. You simply don't have to require 100% "proof" to make an audio purchase. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Nonsense, Harry. ABX has been used to detect the difference between .wav source and 320 kbps mp3s -- which are *extremely* difficult to tell apart. Sensitivity is certainly reduced for comparisons (sighted or blind) when music is used RATHER THAN TEST TONES. This is not a function of sighted versus blind, and the Greenhill tests do not say otherwise. I didn't say it did. What I said was that the Greenhill test clearly shows that the test is sensitive to differences in levels and in white noise, while remaining insensitve with choral music as the source. It says nothing about other tests, pro or con. But it does have relevance for ABX. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. And others will claim they can hear it. Perhaps an ABX test under this circumstance with some training might be of some value. But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. The main value of a difference test is where it produces a null (and by null, it must mean residual levels below the practical or theoretical limits of human hearing). There, subjectivists would have to invoke some new sort of audible effect, akin to homeopathy, where vanishingly small amounts of 'medicine' are said to effect cures. I wouldn't put it past them. It must be nice to be so certain. In the Pro Audio Digest thread of March that I read at J. Junes suggestion last night, one of the correspondents (screened for participation by highly regarded practical professional engineers) antecdotally told of a blind test whereby a friend with a very high level of success could pick out two identical samples (that met the null test and proved bit-identical) on two different brands of gold-plated CD disks. He wasn't challenged. Moreover, the general consensus of the group (which included Jim Johnson and Dan Lavry) was that the CD cutoff was too low and artifacts were often audible as a result, including pre-ringing and/or phase shift, and that 64K was the necessary minimum to avoid even the possibility of problems. Please note that this directly contradicts Arny's recent assertions here that CD's are audibly at the level of ultimate transparency(1), and that the 66khz/20bit recommendation of the Japanese hi-rez group in the mid-90's was nothing but marketing-driven propoganda(2). What I took away after following the discussion, which was one of two major topics for the month, was that real engineers are very aware of what is NOT known (including standards of audibility and a means of simulating what the ear really can hear)...and that the folks here and elsewhere who are so sure they know the truth are not real scientists or engineers. On the other hand, some of us having been saying that for years and shouldn't be surprised. But it is nice to have the real pros reinforce the opinion. Footnotes: (1) A. Krueger, RAHE, "Cable Upgrade Solutions", March 26: "The CD format has always and continues to prvoide sonically transparent recording and reproduction of music....a format that is far from perfect can be sonically transparent." (2) A. Krueger, RAHE, "Need Vinyl to Digital Advice", March 8 (in reply to HL): "HL:The Japanese consortium that investigated hi-rez in the mid '90's concluded that 66khz/20 bits was the minimum beyond which differences could not be heard. "AK: Poor quality tests by advocates of a now-failed technology does not establish a scientific fact. " |
#10
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Arny Krueger" wrote in message
... "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, * This means to determine which bit-rate of MP3 or other codec to choose. Very hi-fi. and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. * Yeah, take that amplifier and shove it into your PC. Ditto the tuner. Maybe the CD player. Wonderful way to listen to components. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. * If such difference is strong enough to overcome PC digitalization. There ain't no such thing. The world is full of differences that are measurable but not audible at all. So, one caveat is that the difference has to be audible, and not all differences are audible. Secondly, the audibility of many differences are contingent on a wide variety of influences, two of the stronger ones being the sensitivity of the listener, and some properties of the music being used for the comparison that may be non-obvioius. * Who are you speaking to? Not me. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Or, you picked the wrong music or associated components. Or, the listener's sensitivity was atypically poor for any number of different reasons. * Who are you speaking to? Not me. Unfortunately, this is not true. So far so good. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. Ironically, these same listeners would no doubt report a plethora of audible differences in a sighted evaluation. ;-) * Straw man supposition, at best. Flame at worst. The test is as much a test of the listener as it is of the items under test. Name a subjective test methodology where this *isn't* true. Some people like to play a silly game where they assert that a certain general problem that afflicts just about any test is a problem of just their least favorite kind of listening evaluation. * The point is: if you want to choose hi-fi, you listen yourself. If you want a scientific test, you prescreen people for sensitivity. You don't use ABX willy-nilly at random....you do a carefully constructed, controlled scientific test, or you just use your ears and do your best. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. Tell that to the ITU standards committee who state otherwise. where easily identifiable distortions can be used at various levels of impact to "train" listeners, Surely true for more than just codecs, so this is also a false claim. Okay.....train for "natural timbre" across a wide variety of instruments. who then listen to the blind samples but "knowing" what they are listening for. Which opens yet another general problem with subjective tests. Differences are very often easier to detect if you know what to listen for, so how do you know what to listen for until someone has first actually heard it? You don't ... that's partly the point. Open-ended evaluation is called that because you don't start with preconceptions or "knowing" differences...you listen for some time on a wide variety of music, form some hypothesis viz a viz differences when those differences crystalize via your ear/brain synthesis, and then switch to synched A-B comparisons across some sections of music that get at it, and then move on to another audio manifestation that has revealed itself to you. When you are done, you've concluded an audio profile of the component you are switching. Such a profile may be (usually is) somewhat multi-dimensional. Open-end evaluation of audio gear does not work this way. Actually, open-end evaluation doesn't work at all. It is open-ended all right - it contains poor-to-non-existent safeguards against false positives. Yes, but it also doesn not result in false negatives. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. A truism that all by itself can't support any particular viewpoint. Wrong...it works directly against ABX and it's need for "known distortion or flaw" and "staged training". That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Fact is that lots of people report fatigue while listening by *any* means that efficiently rejects false positives. The reason is pretty obvious, most of their former experiences were all about false positives. Take those away and they suddenly have to produce more than fond wishes that they heard a difference. Well, that is the objectivist hypothesis. Proof that it is correct? Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. There ain't no such rule. Evidence that you don't really know that much about test design. Absolutely there is such a rule. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. This is one of those statements that immediatly raises a red flag. Broadband white noise, being an unweighted varying and random combination of all frequencies, is very much immune audible effects due to frequency shifts. Theresfore it is a nonsense statement because it is contingent on something that doesn't ever happen. Okay, I should have said "frequency shifts in white noise reproduction". We are (or at least I am) talking about open-ended evaluation of audio components, you know. As soon as it comes to listening to music, sensitivity decreases or disappears. Name a subjective test methodology where this *isn't* true. Long term listening. This isn't just speculation No, its mostly false and mistleading claims, that shouldn't even qualify as reasonable speculation. Where is your "IMO". Hardly a fact. ....review the Greenhill tests in Stereo Review (search index at their site). Greenhill was an advocate of ABX at the time those tests were written and published. I spoke with him personally for hours in those days. Great for you. Your point? Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. Been there, done that many times. The problem with this methodology is that it is easily overwhelmed by trivial differences such as phase shift of any kind but linear phase. You end up with two waves that sound very much alike, but generate relatively large difference signals. The irrelevant information in the difference signal masks the difference that was being listened for. Well, maybe Acel just listens to sine waves. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. This can easily happen, because test equipment has become very, very sensitive. And others will claim they can hear it. You may hear a large difference signal, but is it all that relevant? Perhaps an ABX test under this circumstance with some training might be of some value. You can learn how to do that at www.pcabx.com . But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. www.pcabx.com has links to everything you need to do high quality listening tests as easily as possible, given the inherent problems with listening tests in general, such as the ones I mentioned above. It would have helped very much if you had responded either to me or to Bill Eckle, but not both at once. |
#11
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Steven Sullivan" wrote in message
Arny Krueger wrote: "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. There ain't no such thing. The world is full of differences that are measurable but not audible at all. So, one caveat is that the difference has to be audible, and not all differences are audible. Secondly, the audibility of many differences are contingent on a wide variety of influences, two of the stronger ones being the sensitivity of the listener, and some properties of the music being used for the comparison that may be non-obvioius. However, it only takes one positive ABX test run to demonstrate that the person who took the test could differentiate the sound of A and B. (Assuming of course that the test was set up properly.) You have to be careful. Every once in a while people win on 100:1 shots. And it only takes one person demonstrably hearing a difference, to 'prove' that the two things sound different. That's not the same as saying 'anyone' will be able to hear the difference, of course. We never found any "golden ears" with ABX. We did find tin ears. We did find people who got lucky, but couldn't duplicate their short-run results with longer runs. |
#12
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does it say about the average music listener's ability to correctly identify difference in a sighted comparison? There, one is compounding a lack of training, with a surfeit of bias. Yet that's the regimen used most commonly in audio equipment 'reviewing'. Any wonder that the 'audiophile' is something of a joke? |
#13
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 30, 11:28 am, "Harry Lavo" wrote:
"bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? Nobody *needs* training to take an ABX test. The purpose of training is to heighten the subject's ability to hear the difference being tested. An ABX test tells you what you can hear right now, whether you've had training or not. In particular, if someone claims he can hear a difference between A and B, then he should need no training at all in order to demonstrate that ability in a blind test. That he usually can't do so merely proves he's a phony. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, .This is also false Wishful thinking. That ABX was developed to test codecs? where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. Your whole paragraph was pseudoscience, not just the last bit of nonsense. You're making everything up as you go along, Harry. As for subjects dropping out from "listening fatigue," I can cite you numerous published studies where this did not happen. Can you cite any where it did? Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. More evidence (admittedly anecdotal) than has been presented to validate that it works for open-ended evaluation of audio components. You don't even know what "open-ended evaluation of audio components" means. It's just a nonsense phrase you throw into every post because you haven't any real arguments. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. To paraphrase, you haven't a shred of evidence that long term, exploratory tests paired with short term comparisons doesn't overcome this limitation. We are talking MUSIC, after all...not white noise. We are talking masking. Something else you appear to know nothing about. Do you honestly believe that masking doesn't happen when you listen to music, Harry? Yes, I believe you do. bob |
#14
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"Steven Sullivan" wrote in message ... Harry Lavo wrote: "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. ....and then,once these 'differences' have been 'heard' to the listener's satisfaction, it's time to see if they're real...via a blind comparison. You just can't get around that. Sure you can. You simply don't have to require 100% "proof" to make an audio purchase. Now you're moving the goalposts. I'm talking about establishing audible difference, period, not making purchasing decisions. There was nothing about 'purchases' in the stuff you quoted. For the 10,000th time, no one says anyone 'has' to have 100% proof of audible difference *to make a purchase*. But when people ask (perhaps in relation to a purchase, perhaps not) whether one piece of gear SOUNDS 'better' than another (implying difference), or assert same, they are making an assumption that either stands or falls based on evidence. And 'evidence' from sighted comparison is inevitably prone to bias, whether the comparison is 'open-ended' or not. If you're happy with that, fine, jsut don't act like you've got more than '50%' proof for your claim. If you do want to determine if *you've* heard a *real* difference, then you cannot get around the need for a blind comparison. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Nonsense, Harry. ABX has been used to detect the difference between .wav source and 320 kbps mp3s -- which are *extremely* difficult to tell apart. Sensitivity is certainly reduced for comparisons (sighted or blind) when music is used RATHER THAN TEST TONES. This is not a function of sighted versus blind, and the Greenhill tests do not say otherwise. I didn't say it did. What I said was that the Greenhill test clearly shows that the test is sensitive to differences in levels and in white noise, while remaining insensitve with choral music as the source. It says nothing about other tests, pro or con. But it does have relevance for ABX. All audio comparisons, sighted or not, are affected by significant changes in level. And it is well-known from psychoacoustics that test signals can reveal differences that are masked by more complex signals. So why single out ABX? These effects operate generally and are well-known to people who advocate scientific standards of 'audio reviewing'. If there is ignorance of these factors, it's more likely to be on the part of the sighted-paradigm advocates. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. I think others will argue that it is likely you will be able to meausure some difference, but you won't be able to hear it. And others will claim they can hear it. Perhaps an ABX test under this circumstance with some training might be of some value. But to do all this in a well controlled test is, again, a research project and not an amateur, home-oriented test. The main value of a difference test is where it produces a null (and by null, it must mean residual levels below the practical or theoretical limits of human hearing). There, subjectivists would have to invoke some new sort of audible effect, akin to homeopathy, where vanishingly small amounts of 'medicine' are said to effect cures. I wouldn't put it past them. It must be nice to be so certain. In the Pro Audio Digest thread of March that I read at J. Junes suggestion last night, one of the correspondents (screened for participation by highly regarded practical professional engineers) antecdotally told of a blind test whereby a friend with a very high level of success could pick out two identical samples (that met the null test and proved bit-identical) on two different brands of gold-plated CD disks. He wasn't challenged. Perhaps he should have been. But then again, what's one more anecdotal report, no doubt lacking details of test conditions, number of trials, statistics? I've seen 'trained professionals' in audio and in engineering (as well as 'audio engineers) make some outright nonsensical claims before. Moreover, the general consensus of the group (which included Jim Johnson and Dan Lavry) was that the CD cutoff was too low and artifacts were often audible as a result, including pre-ringing and/or phase shift, and that 64K was the necessary minimum to avoid even the possibility of problems. Doesn't sound like an *inherently* audible result of the technology to me. I'm sure both Lavry and JJ are aware that technology can be implemented well, or suboptimally. Please note that this directly contradicts Arny's recent assertions here that CD's are audibly at the level of ultimate transparency(1), and that the 66khz/20bit recommendation of the Japanese hi-rez group in the mid-90's was nothing but marketing-driven propoganda(2). No, it doesn't contradict (1), as long as JJ and Lavry used a qualifier like 'often' and Arny didn't use the qualifier 'always' to mean that 'a CD is always transparent to its source'. Which it appears he didn't. Do Lavry and JJ claim that 16/44 simply *cannot* be transparent to source? What I took away after following the discussion, which was one of two major topics for the month, was that real engineers are very aware of what is NOT known (including standards of audibility and a means of simulating what the ear really can hear)...and that the folks here and elsewhere who are so sure they know the truth are not real scientists or engineers. On the other hand, some of us having been saying that for years and shouldn't be surprised. But it is nice to have the real pros reinforce the opinion. Hilarious that you're now becoming a fan of Lavry and JJ. I've been reading their stuff for some years now. Btw, how did you access the Pro list? I couldn't find it. ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#15
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"Arny Krueger" wrote in message ... "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, * This means to determine which bit-rate of MP3 or other codec to choose. Very hi-fi. Have you ever tried to tell a well-encoded high-bitrate MP3 from source? and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. * Yeah, take that amplifier and shove it into your PC. Ditto the tuner. Maybe the CD player. Wonderful way to listen to components. I suspect Arny is referring to ABX tests generally, not ABX software -- the latter is a tool for comparing sound files. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. * If such difference is strong enough to overcome PC digitalization. 'PC digitalization' being different from 'digitalization' exactly how? Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. Tell that to the ITU standards committee who state otherwise. ITU claims ABX was developed to test audio codecs?? How are they defining 'codec'? You don't ... that's partly the point. Open-ended evaluation is called that because you don't start with preconceptions or "knowing" differences... Of course you do, Harry -- that's what bias is. These 'preconceptions' maybe be conscious, or not. But they're there. Just seeing that two piece of gear *look* different, can be enough to induce 'preconceptions' of audible difference. Just 'knowing' that you are going to listen to more than one thing, is enough. And around and around we go.... ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#16
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Arny Krueger wrote:
"Steven Sullivan" wrote in message Arny Krueger wrote: "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. There ain't no such thing. The world is full of differences that are measurable but not audible at all. So, one caveat is that the difference has to be audible, and not all differences are audible. Secondly, the audibility of many differences are contingent on a wide variety of influences, two of the stronger ones being the sensitivity of the listener, and some properties of the music being used for the comparison that may be non-obvioius. However, it only takes one positive ABX test run to demonstrate that the person who took the test could differentiate the sound of A and B. (Assuming of course that the test was set up properly.) You have to be careful. Every once in a while people win on 100:1 shots. True. There's no 'absolute' proof. But if the one person did enough trials, the chance of a 'miracle' is thereby reduced. I include a large trial number in 'proper set up', if we're going to put all our bets on one person. And it only takes one person demonstrably hearing a difference, to 'prove' that the two things sound different. That's not the same as saying 'anyone' will be able to hear the difference, of course. We never found any "golden ears" with ABX. We did find tin ears. We did find people who got lucky, but couldn't duplicate their short-run results with longer runs. Ah, probability.... ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#17
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
bob wrote:
On Mar 30, 11:28 am, "Harry Lavo" wrote: "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? Nobody *needs* training to take an ABX test. The purpose of training is to heighten the subject's ability to hear the difference being tested. An ABX test tells you what you can hear right now, whether you've had training or not. In particular, if someone claims he can hear a difference between A and B, then he should need no training at all in order to demonstrate that ability in a blind test. That he usually can't do so merely proves he's a phony. Well, a phony knows he's lying. Assuming sincerity, I'd say it proves he's 'probably wrong' at best, or 'probably deluded' at worst. ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#18
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Getting back to the original article, isn't this similar to the Bob
Carver "trick" he used a long time ago when he attempted to match the "sound" of his solid state amplifier to a C-J tube amp for the boys at Stereophile, and then a Levinson ML-2 in a production run of one model of his amplifiers? I'd have to investigate, but it kind of seems similar in a way. mp |
#19
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
|
#20
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 30, 11:32 am, "Harry Lavo" wrote:
I didn't say it did. What I said was that the Greenhill test clearly shows that the test is sensitive to differences in levels and in white noise, while remaining insensitve with choral music as the source. It says nothing about other tests, pro or con. But it does have relevance for ABX. Denial of the existence of masking noted. snip It must be nice to be so certain. In the Pro Audio Digest thread of March that I read at J. Junes suggestion last night, one of the correspondents (screened for participation by highly regarded practical professional engineers) antecdotally told of a blind test whereby a friend with a very high level of success could pick out two identical samples (that met the null test and proved bit-identical) on two different brands of gold-plated CD disks. He wasn't challenged. And I notice you're not challenging him, either. Moreover, the general consensus of the group (which included Jim Johnson and Dan Lavry) was that the CD cutoff was too low and artifacts were often audible as a result, including pre-ringing and/or phase shift, and that 64K was the necessary minimum to avoid even the possibility of problems. Please note that this directly contradicts Arny's recent assertions here that CD's are audibly at the level of ultimate transparency(1), and that the 66khz/20bit recommendation of the Japanese hi-rez group in the mid-90's was nothing but marketing-driven propoganda(2). Yes, it's long been recognized by a lot of technical types that 16/44.1 might not be quite enough for perfect transparency. And how did they figure this out? Blind tests, Harry. What I took away after following the discussion, which was one of two major topics for the month, was that real engineers are very aware of what is NOT known (including standards of audibility and a means of simulating what the ear really can hear)...and that the folks here and elsewhere who are so sure they know the truth are not real scientists or engineers. On the other hand, some of us having been saying that for years and shouldn't be surprised. But it is nice to have the real pros reinforce the opinion. Harry slays his straw man. Well, not really. Harry slays his straw man while citing the same tests he claims his straw men are erroneously citing. Harry really ought to try to get his story straight. bob |
#21
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
William Eckle wrote:
The classic abx listening test is a chore to setup and perform. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. Substitute some element for another which is thought to possibly make an audible difference and if no difference can be heard then the claim of amp or wire etc. differences are moot and any subjective claims highly questionable. Here is a new way to perform the same kind of test with greatly simplified methods that create the context where a possible difference can be shown to exist by listening alone and also scientifically valid. http://theaudiocritic.com/blog/index...Id=35&blogId=1 Don't miss the link at the end of this short article that describes the software and methods used. This is in no way a substitute for a valid listening test. It is a measurement, a technical curiosity and no more. A given result in this test may or may not be audible, the answer to which can only be determined by... a listening test! Gary Eickmeier |
#22
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Steven Sullivan" wrote in message
... Harry Lavo wrote: "Arny Krueger" wrote in message ... "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, * This means to determine which bit-rate of MP3 or other codec to choose. Very hi-fi. Have you ever tried to tell a well-encoded high-bitrate MP3 from source? I have listened to 192k MP3 on the main system on which I normally play the CD's, and after five minutes my ears bled. and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. * Yeah, take that amplifier and shove it into your PC. Ditto the tuner. Maybe the CD player. Wonderful way to listen to components. I suspect Arny is referring to ABX tests generally, not ABX software -- the latter is a tool for comparing sound files. His exact quote in it's totality, as above: AK: Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Sounds like a PC test to me. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. * If such difference is strong enough to overcome PC digitalization. 'PC digitalization' being different from 'digitalization' exactly how? Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. Tell that to the ITU standards committee who state otherwise. ITU claims ABX was developed to test audio codecs?? How are they defining 'codec'? It was developed to help test telephone transmission techniques, which were early forerunners of what we nowadays call audio codecs. You don't ... that's partly the point. Open-ended evaluation is called that because you don't start with preconceptions or "knowing" differences... Of course you do, Harry -- that's what bias is. These 'preconceptions' maybe be conscious, or not. But they're there. Just seeing that two piece of gear *look* different, can be enough to induce 'preconceptions' of audible difference. Just 'knowing' that you are going to listen to more than one thing, is enough. You miss my point completely. Whether deliberately or not I do not know. Perhaps if you hadn't cut two-thirds of it out..... And around and around we go.... And up and down go the horses.... |
#23
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"bob" wrote in message
... On Mar 30, 11:28 am, "Harry Lavo" wrote: "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? Nobody *needs* training to take an ABX test. The purpose of training is to heighten the subject's ability to hear the difference being tested. An ABX test tells you what you can hear right now, whether you've had training or not. In particular, if someone claims he can hear a difference between A and B, then he should need no training at all in order to demonstrate that ability in a blind test. That he usually can't do so merely proves he's a phony. You ignore the fact that Harmon rejects over 40% of people as simply unable to be consistent in the use of an ABX test. Since obviously 100% of people can listen to two components playing music, this means that USING THE ABX TEST many people are inadequate to use the test. It doesn't mean they are phonys. Nor does it mean they couldn't hear a difference using other tests. Or using no tests. All you can conclude is that USING THE ABX TEST, they can hear no difference with any consistency. Even when there is one. So in using the test without knowing if there is a difference, if one "flunks" the test, then one does not know whether that is because there is no difference, or because one is not good at ABX'ng. And this cannot be determined without pre-screening. You oversimplifications undermine any legitimacy to your argument. The test is as much a test of the listener as it is of the items under test. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, .This is also false Wishful thinking. That ABX was developed to test codecs? Telephone transmission compression, yes. A forerunner. where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. Your whole paragraph was pseudoscience, not just the last bit of nonsense. You're making everything up as you go along, Harry. As for subjects dropping out from "listening fatigue," I can cite you numerous published studies where this did not happen. Can you cite any where it did? Anecdotally, in several efforts at testing here on usenet (not in this forum) by people actively attempting to use the technique (not subjectivists), yes. Can I cite the posts, no....some of them were seven-eight years ago. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. More evidence (admittedly anecdotal) than has been presented to validate that it works for open-ended evaluation of audio components. You don't even know what "open-ended evaluation of audio components" means. It's just a nonsense phrase you throw into every post because you haven't any real arguments. That is pure baloney. I have defined it over and over and have done so just recently. You just don't want to acknowledge what it is, for there is no way it can be accomodated in an ABX test. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. To paraphrase, you haven't a shred of evidence that long term, exploratory tests paired with short term comparisons doesn't overcome this limitation. We are talking MUSIC, after all...not white noise. We are talking masking. Something else you appear to know nothing about. Do you honestly believe that masking doesn't happen when you listen to music, Harry? Yes, I believe you do. I believe masking happens when listening for differences in an ABX test. I also believe "sharpening" happens when listening in extended listening tests, if one is a careful listener. Music is not just "sound", a fact that you and others ignore time and time again, at the expense of undermining your own arguments. |
#24
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Steven Sullivan" wrote in message
... Harry Lavo wrote: "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does it say about the average music listener's ability to correctly identify difference in a sighted comparison? There, one is compounding a lack of training, with a surfeit of bias. What it does is rely on the listeners ability to judge the natural sound of music. If the person is good at it, he will probably reach a valid conclusion. If he has a tin ear, he probably will reach a "no difference" or "random difference" conclusion, both of which are okay for him. But he will judge based on music heard in a more natural way without any need to be "trained". And he can listen multidimensionally, rather than as a test subject. Yet that's the regimen used most commonly in audio equipment 'reviewing'. Any wonder that the 'audiophile' is something of a joke? And any wonder why the attempts to impose sterile and artificial tests are mostly just simply ignored? |
#25
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Harry Lavo" wrote in message
"bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does this say about anybody's ability to reliably hear differences with or without training? As usual we see a post that tries to blame the problems of the world on ABX. One core problem is the horrific propensity of sighted evaluations to product false positives. If the number of positive results that is used to judge is based on a listening methodolgy that spawns false positives like pregnant pacific salmon produce eggs, what can compete? The test is as much a test of the listener as it is of the items under test. True of any listening test involving subtle differences, or even differences that aren't glaringly obvious. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, .This is also false Wishful thinking. No factually wrong. ABX was developed and popularized before testing codecs was a serious business. Back in the middle 1970s, MP3 was 20 or more years in the future. where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. Especially given that "open-ended" listening is whatever Harry wants it to be today. This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. No foundation has been laid for this claim. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. More evidence (admittedly anecdotal) than has been presented to validate that it works for open-ended evaluation of audio components. Keys words "admittedly anecdotal". IOW there's nothing but unsubstantiated stories to back it up. As other posters have said, urban legends, but in this case urban legends know to only one person. ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. To paraphrase, you haven't a shred of evidence that long term, exploratory tests paired with short term comparisons doesn't overcome this limitation. We are talking MUSIC, after all...not white noise. What is an exploratory test? I have an idea of what an exploratory test is, and it no way does it conflict with the use of ABX, ABC/hr or whatever. This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Where you won't find it. But if anyone wants to read what Greenhill actually found (rather than someone's re-invention of it), e-mail me and I can send you the article. Agreed, I believe I have the text of those tests on my hard drive. And if you wish to email me I can send you an accurate and complete Excel table of the results. Seems like rehashing just one test done about 20 years ago is pretty senseless, anyhow. Harry is peddling pure pseudoscience here. You've heard from a true ABX believer. Who might that be? Ask for the validation test that this technique, developed very specifically for codec distortions, works as the best tool for open-ended evaluation of audio components. This is nonsense. ABX predates codec tests by about 20 years. I won't say that there weren't codecs way back then, but the ones that provided meaninful amount of compression had so many audible faults that they were only proposed for use with telephones, not hifis. |
#26
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Harry Lavo" wrote in message
"Steven Sullivan" wrote in message ... Harry Lavo wrote: "Arny Krueger" wrote in message ... "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, * This means to determine which bit-rate of MP3 or other codec to choose. Very hi-fi. Have you ever tried to tell a well-encoded high-bitrate MP3 from source? I have listened to 192k MP3 on the main system on which I normally play the CD's, and after five minutes my ears bled. The answer would be thus be "no", as the question presumes bias controls, and none are in evidence. A good 192K codec can be suitable challenge for a listener, once the classic prejudices of audiophilia are kept under control. Let Harry provide a CD track of his choosing. It will be returned to him repeated 16 times, some repetitions precise copies, and some repetitions round-tripped through 196 kB MP3-land. Harry will tell us which are the "ear bleed" tracks. and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a carefuljob of comparing components using one of the house's systems. * Yeah, take that amplifier and shove it into your PC. A ludicrous and dismissive comment. Ditto the tuner. Maybe the CD player. Wonderful way to listen to components. Proof? I suspect Arny is referring to ABX tests generally, not ABX software -- the latter is a tool for comparing sound files. His exact quote in it's totality, as above: AK: Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Sounds like a PC test to me. PC's are widely used for audio production purposes. Perhaps Harry has a "ear bleed" remark to go with all of the recordings he already has that were unbeknownst to him, produced on a PC? ;-) Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. * If such difference is strong enough to overcome PC digitalization. 'PC digitalization' being different from 'digitalization' exactly how? Harry has no response - he must be agreeing with the obvious point that there need be no difference between PC digitization and other forms of digitization. Interesting that Harry chooses a word whose primary meaning is: "To administer digitalis in a dosage sufficient to achieve the maximum therapeutic effect without producing toxic symptoms." Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. Tell that to the ITU standards committee who state otherwise. ITU claims ABX was developed to test audio codecs?? How are they defining 'codec'? It was developed to help test telephone transmission techniques, which were early forerunners of what we nowadays call audio codecs. I don't know about this. You don't ... that's partly the point. Open-ended evaluation is called that because you don't start with preconceptions or "knowing" differences... Of course you do, Harry -- that's what bias is. These 'preconceptions' maybe be conscious, or not. But they're there. Just seeing that two piece of gear *look* different, can be enough to induce 'preconceptions' of audible difference. Just 'knowing' that you are going to listen to more than one thing, is enough. Harry also did not respond to this in any meaninful way. |
#27
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Gary Eickmeier wrote:
This is in no way a substitute for a valid listening test. It is a measurement, a technical curiosity and no more. A given result in this test may or may not be audible, the answer to which can only be determined by... a listening test! What you say is true, strictly speaking. You are just describing the idea that all empirical phenomenon are presumed to be contingent. But practically it is not so relevant. I can use a sort of paraphrase that actually came from The Audio Critic many years ago. The idea was that if you wanted to investigate fast animals you'd probably want to check out the gazelle, a couple of the big predator cats, a thoroughbred race horse and so forth. You'd probably not waste your time investigating a pig, or clocking an opossum. On the other hand, if someone credible told you that they'd observed a fast pig, it might be a different matter and out of curiosity you might want to check it out. It is the same here. With this test, if the measurements show no (or very little) difference, and upon further investigation by listening you then conclude that whatever differences you actually measured are indistinguishable in your listening test, then you can rightly conclude and have grounds for certainty that in the future audible differences are not going to manifest below this threshold. It is just reasoning by induction. mp |
#28
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 31, 10:23 am, "Harry Lavo" wrote:
You ignore the fact that Harmon rejects over 40% of people as simply unable to be consistent in the use of an ABX test. Well, let's see a reference for this, please. Harman certainly screened lots of people out, but not because they were "unable" to "use" any test. Harman screened them out because their hearing wasn't good enough for the research Harman was doing. If Harman used ABX tests at all, it was because they know ABX tests WORK. Your tendentious misinterpretation of Harman's work is astounding. Let's assume, just for the sake of argument, that there really are people who are "unable to use ABX" (as opposed to "unable to get the results Harry wants"). Why would Harman screen them out? Harman wasn't using ABX tests in its research. Why would it care whether people could "use" them or not? snip I believe masking happens when listening for differences in an ABX test. I also believe "sharpening" happens when listening in extended listening tests, if one is a careful listener. Music is not just "sound", a fact that you and others ignore time and time again, at the expense of undermining your own arguments. Here in the reality-based community, Harry, masking happens all the time, no matter how you're listening. But in order to hold on to your baseless beliefs about ABX tests, you have to deny that masking occurs when listening to music. That's how far out of the scientific world you've stepped, Harry. Your arguments resemble those of the Creationist who has to deny carbon dating and genetic mutation. bob |
#29
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"Steven Sullivan" wrote in message ... Harry Lavo wrote: "Arny Krueger" wrote in message ... "Harry Lavo" wrote in message "William Eckle" wrote in message om... The classic abx listening test is a chore to setup and perform. Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, * This means to determine which bit-rate of MP3 or other codec to choose. Very hi-fi. Have you ever tried to tell a well-encoded high-bitrate MP3 from source? I have listened to 192k MP3 on the main system on which I normally play the CD's, and after five minutes my ears bled. And of course, you weren't listening blind, right? If you were, that must've been an extremely poorly made MP3 then. I doubt you could tell the ones I make, apart from source. and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. * Yeah, take that amplifier and shove it into your PC. Ditto the tuner. Maybe the CD player. Wonderful way to listen to components. I suspect Arny is referring to ABX tests generally, not ABX software -- the latter is a tool for comparing sound files. His exact quote in it's totality, as above: AK: Depends. PCABX is an exact test for many kinds of listening tests that people are interested in today, and a useful approximation for a wide variety of tests. It's easier than going to a store and doing a careful job of comparing components using one of the house's systems. Sounds like a PC test to me. True, I don't see how one can easily compare components using PCABX, *if* they aren't among the data set Arny has provided. Its clear strength is the ability to determine if a difference, any difference, can be shown to exist by listening alone in a scientifically valid way. * If such difference is strong enough to overcome PC digitalization. 'PC digitalization' being different from 'digitalization' exactly how? Simply not true of ABX, since ABX was developed and popularized before the audio world even knew what a codec in the modern sense was. Tell that to the ITU standards committee who state otherwise. ITU claims ABX was developed to test audio codecs?? How are they defining 'codec'? It was developed to help test telephone transmission techniques, which were early forerunners of what we nowadays call audio codecs. You don't say? ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#30
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
Harry Lavo wrote:
"Steven Sullivan" wrote in message ... Harry Lavo wrote: "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does it say about the average music listener's ability to correctly identify difference in a sighted comparison? There, one is compounding a lack of training, with a surfeit of bias. What it does is rely on the listeners ability to judge the natural sound of music. If the person is good at it, he will probably reach a valid conclusion. If he has a tin ear, he probably will reach a "no difference" or "random difference" conclusion, both of which are okay for him. But he will judge based on music heard in a more natural way without any need to be "trained". And he can listen multidimensionally, rather than as a test subject. You argument presumes where it needs to proved Yet that's the regimen used most commonly in audio equipment 'reviewing'. Any wonder that the 'audiophile' is something of a joke? And any wonder why the attempts to impose sterile and artificial tests are mostly just simply ignored? The only place they should be 'imposed' is in the context of professionally reviewing the sound...where the stated intent is to inform the consumer (though the actual intent may be to sell advertising space). That they are 'ignored' there is a testament to the bankruptcy of professional 'audiophile' journalism. ('Ignored' is in quotes because in fact, magazines like Stereophile actually print anti-DBT screeds). ___ -S "As human beings, we understand the world through simile, analogy, metaphor, narrative and, sometimes, claymation." - B. Mason |
#31
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Arny Krueger" wrote in message
... "Harry Lavo" wrote in message "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does this say about anybody's ability to reliably hear differences with or without training? It doesn't say anything about other tests...what it does say is 40%+ of people have difficulty getting a valid result using ABX even when there are known subtle differences. As usual we see a post that tries to blame the problems of the world on ABX. One core problem is the horrific propensity of sighted evaluations to product false positives. If the number of positive results that is used to judge is based on a listening methodolgy that spawns false positives like pregnant pacific salmon produce eggs, what can compete? The test is as much a test of the listener as it is of the items under test. True of any listening test involving subtle differences, or even differences that aren't glaringly obvious. Wrong for two reasons. First, it is a particularly bad problem with ABX because that difficulty creates a false obstacle...eg. in Greenhills test, suppose 40% of the testers who heard no difference had been thrown out as incompetent testers. The results among those remaining would have looked less like chance, wouldn't they? Many other tests don't require throwing out testers to be valid...they have different ways of dealing with the issue. Second, the problem is test specific. It is quite possible that those people would have no problem doing a blind A-B preference test . Or doing a monadic or protomonadic semantic scaling test. This gets at the stress factor that is seemingly inherent in substituting an identification test for a preference test or a "liking test" when it comes to music. And this in turn gets at the fact that identification tests per se reduce emotional involvement to zip. That's one reason ABC/hr is a better technique than ABX, although it is still a quasi-id test. It at least allows the sound of the real world to be a factor. Moreover, both ABX and ABC/hr were developed and optimized specifically for codec testing, .This is also false Wishful thinking. No factually wrong. ABX was developed and popularized before testing codecs was a serious business. Back in the middle 1970s, MP3 was 20 or more years in the future. Exactly...and compressed telephone transmissions were the forerunner of today's codecs. In fact, as you well know, JJ moved from the former to the latter. where easily identifiable distortions can be used at various levels of impact to "train" listeners, who then listen to the blind samples but "knowing" what they are listening for. Open-end evaluation of audio gear does not work this way. The human brain tries to relate the sound to a real sound, and doesn't even know "what" to listen for...timbre, soundspace, subtles distortions, etc. That seems to be why the result of open-ended listening via ABX results in almost immediate listening fatique...it is a total unnatural use of the technique for this purpose. This is idle speculation by someone who knows nothing about the subject he's talking about. There isn't a shred of real evidence for any of it. It's pseudoscience. Especially given that "open-ended" listening is whatever Harry wants it to be today. No, open-ended listening has been repeatedly defined here by me. And you know it...there probably are a half dozen posters to this forum who could paraphrase it. This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. No foundation has been laid for this claim. See below. Put simply it violates the first cardinal principal of test design...that is to prevent any aspect of the test from intervening as a variable. Again, you haven't a shred of evidence that ABX or ABC/hr tests interfere with perception in any way. More evidence (admittedly anecdotal) than has been presented to validate that it works for open-ended evaluation of audio components. Keys words "admittedly anecdotal". IOW there's nothing but unsubstantiated stories to back it up. As other posters have said, urban legends, but in this case urban legends know to only one person. I seem to recall another poster here who refers to anecdotal evidence when asked to back his claims. Sauce for the Goose, Arny??? (Except you don't admit that is the basis for your evidence until pressed to the wall). ABX can be used for crude audio measures....volume, frequency shifts in white noise, etc. As soon as it comes to listening to music, sensitivity decreases or disappears. It is certainly true that it is easier to hear differences in level and FR using test tones than using music. This has nothing to do with any particular test. It has to do with the way the human hearing mechanism works, and it is true no matter what listening method you use. To paraphrase, you haven't a shred of evidence that long term, exploratory tests paired with short term comparisons doesn't overcome this limitation. We are talking MUSIC, after all...not white noise. What is an exploratory test? I have an idea of what an exploratory test is, and it no way does it conflict with the use of ABX, ABC/hr or whatever. An exploratory test is a test whereby one is trying to get a "fix" on the sonic characteristics of a piece of audio gear, Arny. And the whole ABX, ABC/hr test prtocols are wrong for this sort of thing.....they are not designed for music, and trying to use them for this is simply false "science". This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Where you won't find it. But if anyone wants to read what Greenhill actually found (rather than someone's re-invention of it), e-mail me and I can send you the article. Agreed, I believe I have the text of those tests on my hard drive. Many of us have that test on our hard drives, Arny. Your point?? And if you wish to email me I can send you an accurate and complete Excel table of the results. Seems like rehashing just one test done about 20 years ago is pretty senseless, anyhow. Harry is peddling pure pseudoscience here. You've heard from a true ABX believer. Who might that be? Who am I responding to here, Arny? Hint: it is not you. Ask for the validation test that this technique, developed very specifically for codec distortions, works as the best tool for open-ended evaluation of audio components. This is nonsense. ABX predates codec tests by about 20 years. I won't say that there weren't codecs way back then, but the ones that provided meaninful amount of compression had so many audible faults that they were only proposed for use with telephones, not hifis. Which is exactly when and why ABX was developed/codified by the ITU...and then found not to be so good, so later replaced by ABX/hr for this purpose. But that is hardly finding subtle flaws in a piece of equipment's dynamic reproduction of music, now, is it Arny? |
#32
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Steven Sullivan" wrote in message
... Harry Lavo wrote: "Steven Sullivan" wrote in message ... Harry Lavo wrote: "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does it say about the average music listener's ability to correctly identify difference in a sighted comparison? There, one is compounding a lack of training, with a surfeit of bias. What it does is rely on the listeners ability to judge the natural sound of music. If the person is good at it, he will probably reach a valid conclusion. If he has a tin ear, he probably will reach a "no difference" or "random difference" conclusion, both of which are okay for him. But he will judge based on music heard in a more natural way without any need to be "trained". And he can listen multidimensionally, rather than as a test subject. You argument presumes where it needs to proved Well, let's see how unreasonable or reasonable the presumptions are! 1) "What it does is rely on the listeners ability to judge the natural sound of music.". Now we are judging the quality of audio components to reproduce live music with versimiltude...that is the task. Can you suggest an alternative means of making that judgement that is better? 2) "If the person is good at it, he will probably reach a valid conclusion." This is a truism. If a person is good at judging via the above approach, his judements with regard to the quality of audio components to reproduce live music with versimitude will be good, do you not expect? Would the opposite likely be true? 3) "If he has a tin ear, he probably will reach a "no difference" or "random difference" conclusion, both of which are okay for him. " This needs broken down. Assuming you understand that by "tin ear" I mean he seems to not be able to recognize really good sound from average sound from very poor sound, why would the results be anything but as stated...he either can't hear any difference, or what he hears essentially is random. Do you have a problem with this? 4) "But he will judge based on music heard in a more natural way without any need to be "trained"." In other words, however he listens to music, however carefully or sloppily or backgroundy, however full of discernment or ignorance of live instruments, that is the approach he will bring to open-ended listening. Is there some more likely alternative I have overlooked? No, then let's move on 5) "And he can listen multidimensionally, rather than as a test subject." Meaning, if his mind drifts off to the bass line, and then notices that the snares seem really realistic behind the bass, and that then the alto has "air" around it, and that by golly this is a teriffic reproduction of a jazz trio, he can do so without being asked to focus on "which reproduction of bass sounds like 'X'", a single sound and a single focus...which is one-dimensional. Okay...that's the end of my dissection. What part of the above needs to be proven to be reasonable and likely true? And versus what alternative? Yet that's the regimen used most commonly in audio equipment 'reviewing'. Any wonder that the 'audiophile' is something of a joke? And any wonder why the attempts to impose sterile and artificial tests are mostly just simply ignored? The only place they should be 'imposed' is in the context of professionally reviewing the sound...where the stated intent is to inform the consumer (though the actual intent may be to sell advertising space). A magazine is not a lab. A consumers reports devoted to audio would be out of business in no time...the potential market is simply not large enough. Instead, the magazine reviewers essentially try to relate to the equipment as audiophiles. And audiophiles learn which reviewers judgments square with their own and theirfore which to trust highly (and which to not trust highly). This is how the magazines approach the subject, and this is how audiophiles use and understand the reviews. Given the cost and complexity of a really tight and controlled scientific test, is not this a reasonable approach for a consumer hobbiest magazine to take? Do you suppose this might be so strongly the case that most audio hobby magazines no matter where they originate in the world seem to follow this approach? That they are 'ignored' there is a testament to the bankruptcy of professional 'audiophile' journalism. Or perhaps is a testimony to the financial astuteness of their owners and publishers. ('Ignored' is in quotes because in fact, magazines like Stereophile actually print anti-DBT screeds). One man's screed is another man's alternative take on things. |
#33
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 31, 3:29 pm, "Harry Lavo" wrote:
It doesn't say anything about other tests...what it does say is 40%+ of people have difficulty getting a valid result using ABX even when there are known subtle differences. Still waiting for a reference. snip First, it is a particularly bad problem with ABX because that difficulty creates a false obstacle...eg. in Greenhills test, suppose 40% of the testers who heard no difference had been thrown out as incompetent testers. But there's no evidence that there *were* "incompetent testers" (a category we still haven't established the existence of). The results among those remaining would have looked less like chance, wouldn't they? Given that 5 out of 6 tests had positive outcomes, it's hard to imagine how this could be. (And excluding the 4 worst performers overall from the one negative test wouldn't get you a positive result, either.) Many other tests don't require throwing out testers to be valid...they have different ways of dealing with the issue. Second, the problem is test specific. It is quite possible that those people would have no problem doing a blind A-B preference test . Or doing a monadic or protomonadic semantic scaling test. This gets at the stress factor that is seemingly inherent in substituting an identification test for a preference test or a "liking test" when it comes to music. And this in turn gets at the fact that identification tests per se reduce emotional involvement to zip. The preceding paragraph is entirely empiricism-free. Harry has no evidence that any other test is better than ABX. None. It is baseless belief. That's one reason ABC/hr is a better technique than ABX, although it is still a quasi-id test. ABC/hr ISN'T a better test; it's a different test with a different purpose. You really ought to study up on these things before you write about them, Harry. snip This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. No foundation has been laid for this claim. See below. I looked. Ain't nothin' there. snip An exploratory test is a test whereby one is trying to get a "fix" on the sonic characteristics of a piece of audio gear, Arny. That's not even a test. And, of course, there's nothing to stop someone from doing this "exploration" before taking any test. Indeed, people who claim to hear a difference between cables or amps or whatever have presumably already done this "exploration." Which makes it doubly puzzling why they can't find ANY blind test that will confirm the results of that "exploration." And the whole ABX, ABC/hr test prtocols are wrong for this sort of thing.....they are not designed for music, and trying to use them for this is simply false "science". So if they weren't designed for music, just what source material did the researchers use in the codec tests, Harry? This isn't just speculation....review the Greenhill tests in Stereo Review (search index at their site). Where you won't find it. But if anyone wants to read what Greenhill actually found (rather than someone's re-invention of it), e-mail me and I can send you the article. Agreed, I believe I have the text of those tests on my hard drive. Many of us have that test on our hard drives, Arny. Your point?? And if you wish to email me I can send you an accurate and complete Excel table of the results. Seems like rehashing just one test done about 20 years ago is pretty senseless, anyhow. Harry is peddling pure pseudoscience here. You've heard from a true ABX believer. Who might that be? Who am I responding to here, Arny? Hint: it is not you. Ask for the validation test that this technique, developed very specifically for codec distortions, works as the best tool for open-ended evaluation of audio components. This is nonsense. ABX predates codec tests by about 20 years. I won't say that there weren't codecs way back then, but the ones that provided meaninful amount of compression had so many audible faults that they were only proposed for use with telephones, not hifis. Which is exactly when and why ABX was developed/codified by the ITU...and then found not to be so good, so later replaced by ABX/hr for this purpose. Poppycock. Nobody "replaced" ABX. They developed a variant of ABX for a different purpose. bob |
#34
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Harry Lavo" wrote in message
It must be nice to be so certain. In the Pro Audio Digest thread of March that I read at J. Junes suggestion last night, one of the correspondents (screened for participation by highly regarded practical professional engineers) antecdotally told of a blind test whereby a friend with a very high level of success could pick out two identical samples (that met the null test and proved bit-identical) on two different brands of gold-plated CD disks. He wasn't challenged. That's beause that is a well-known historical problem that has nothing to do with resolution, per se. Moreover, the general consensus of the group (which included Jim Johnson and Dan Lavry) was that the CD cutoff was too low and artifacts were often audible as a result, including pre-ringing and/or phase shift, and that 64K was the necessary minimum to avoid even the possibility of problems. Please note that this directly contradicts Arny's recent assertions here that CD's are audibly at the level of ultimate transparency(1), and that the 66khz/20bit recommendation of the Japanese hi-rez group in the mid-90's was nothing but marketing-driven propoganda(2). Without supporting quotes, this is just another anecdote. BTW, here's some interesting reading: http://www.paudio.com/Pages/presenta...ity/sld001.htm |
#35
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
On Mar 31, 8:20 pm, "Harry Lavo" wrote:
2) "If the person is good at it, he will probably reach a valid conclusion." This is a truism. No, it's actually a falsehood. No matter how good someone is at "open- ended evaluation of audio components" (whatever that is), we know that he will frequently come to demonstrably incorrect conclusions. That's what's wrong with "open-ended evaluation of audio components" (whatever that is): It produces a high incidence of false positives. But, as Steven noted, you are assuming that little problem away. bob |
#36
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Harry Lavo" wrote in message
"Arny Krueger" wrote in message ... "Harry Lavo" wrote in message "bob" wrote in message ... On Mar 28, 11:21 pm, "Harry Lavo" wrote: Unfortunately, this is not true. The researchers at Harman Kardon have found that over 40% of people, even with careful training, cannot reliably distinquish even known differences in their training, and have to be dropped from any testing. This is false. Harmon wasn't even training people to "distinguish differences." All of its subjects could *distinguish* the differences; what they couldn't do was correlate those differences to specific variations in frequency response. That is a much harder task, which is why the failure rate, even after training, was so high. Anyone who's read Sean Olive's work would understand this (assuming they wanted to understand it). So what does this say about the average music listeners ability to use ABX with NO training? What does this say about anybody's ability to reliably hear differences with or without training? It doesn't say anything about other tests...what it does say is 40%+ of people have difficulty getting a valid result using ABX even when there are known subtle differences. If you can't compare ABX to other testing methods, what's your point Harry? |
#37
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
|
#38
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"bob" wrote in message
... On Mar 31, 3:29 pm, "Harry Lavo" wrote: It doesn't say anything about other tests...what it does say is 40%+ of people have difficulty getting a valid result using ABX even when there are known subtle differences. Still waiting for a reference. snip Well, if you are impatient...how about contacting Sean himself? First, it is a particularly bad problem with ABX because that difficulty creates a false obstacle...eg. in Greenhills test, suppose 40% of the testers who heard no difference had been thrown out as incompetent testers. But there's no evidence that there *were* "incompetent testers" (a category we still haven't established the existence of). You certainly can't tell if they are incompetent if you haven't developed a screening protocol to find them. Can you, Bob? The results among those remaining would have looked less like chance, wouldn't they? Given that 5 out of 6 tests had positive outcomes, it's hard to imagine how this could be. (And excluding the 4 worst performers overall from the one negative test wouldn't get you a positive result, either.) There were six tests, only two of which had positive outcomes even summing across all 11 testers. Many other tests don't require throwing out testers to be valid...they have different ways of dealing with the issue. Second, the problem is test specific. It is quite possible that those people would have no problem doing a blind A-B preference test . Or doing a monadic or protomonadic semantic scaling test. This gets at the stress factor that is seemingly inherent in substituting an identification test for a preference test or a "liking test" when it comes to music. And this in turn gets at the fact that identification tests per se reduce emotional involvement to zip. The preceding paragraph is entirely empiricism-free. Harry has no evidence that any other test is better than ABX. None. It is baseless belief. Agreed that it is empiricism-free. But it is not antecdotal-free. As to other tests, where other types of testing have been used, their hasn't been such a plethora of anecdotal-reported problems. So you can view this as an informed opinion by one who spent a good portion of his life helping design and approve tests, albeit in a diferent sensory field. That's one reason ABC/hr is a better technique than ABX, although it is still a quasi-id test. ABC/hr ISN'T a better test; it's a different test with a different purpose. You really ought to study up on these things before you write about them, Harry. It was and is a better test for telephone transmission testing, which is what it was developed for. And it is a better test not on my say so, but because it gives better (e.g. more discriminating) test results regarding the human voice. Because, the ITU says, test subjects have a reference for human voices in the real world. Now, interestingly, the ABX/hr test in some regards represents a semantic differential test (somewhat similar to one of the suggested alternatives for home evaluation of gear) more than it does the original ABX test. snip This is reality, as even people who desire to take the test often drop out before even 15 samples for this very reason. No foundation has been laid for this claim. See below. I looked. Ain't nothin' there. Well it was anecdotal and you have snipped it. So you don't accept anecdotals as even worthy of tentative credibility? snip An exploratory test is a test whereby one is trying to get a "fix" on the sonic characteristics of a piece of audio gear, Arny. That's not even a test. And, of course, there's nothing to stop someone from doing this "exploration" before taking any test. Indeed, people who claim to hear a difference between cables or amps or whatever have presumably already done this "exploration." Which makes it doubly puzzling why they can't find ANY blind test that will confirm the results of that "exploration." No, an exploratory test is where you move swiftly from open-ended listening into short-snippet comparative listening. That is a form of a test...simply not one your approve of. And the whole ABX, ABC/hr test prtocols are wrong for this sort of thing.....they are not designed for music, and trying to use them for this is simply false "science". So if they weren't designed for music, just what source material did the researchers use in the codec tests, Harry? They were listening for specific distortion artifacts...that is what the training is for. For example, here is the sound as it affects cymbal reproduction, at various levels. At what level can you identify it? That is not "listening to music". Not open-ended. And not even close-ended. Instead you have been trained to listen for a specific type of non-musical distortion. snip, to shorten ... Which is exactly when and why ABX was developed/codified by the ITU...and then found not to be so good, so later replaced by ABX/hr for this purpose. Poppycock. Nobody "replaced" ABX. They developed a variant of ABX for a different purpose. They developed it specifically because participants knew what real human voices sounded like and wanted to use that as a frame of reference for evaluation, but ABX didn't allow that. And that test became the standard for telephone transmission testing as a result. Notice any similarities to the reproduction of music, Bob? |
#39
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"Arny Krueger" wrote in message
... "Harry Lavo" wrote in message It must be nice to be so certain. In the Pro Audio Digest thread of March that I read at J. Junes suggestion last night, one of the correspondents (screened for participation by highly regarded practical professional engineers) antecdotally told of a blind test whereby a friend with a very high level of success could pick out two identical samples (that met the null test and proved bit-identical) on two different brands of gold-plated CD disks. He wasn't challenged. That's beause that is a well-known historical problem that has nothing to do with resolution, per se. I see...my guess is this comes as a surprise to most here. Care to elaborate? Moreover, the general consensus of the group (which included Jim Johnson and Dan Lavry) was that the CD cutoff was too low and artifacts were often audible as a result, including pre-ringing and/or phase shift, and that 64K was the necessary minimum to avoid even the possibility of problems. Please note that this directly contradicts Arny's recent assertions here that CD's are audibly at the level of ultimate transparency(1), and that the 66khz/20bit recommendation of the Japanese hi-rez group in the mid-90's was nothing but marketing-driven propoganda(2). Without supporting quotes, this is just another anecdote. Anecdotal here, and never positioned as anything other than a report. Anybody can go read for themselves...just sign up for the digest version and pull the report for the entire month of March and read it. BTW, here's some interesting reading: http://www.paudio.com/Pages/presenta...ity/sld001.htm |
#40
Posted to rec.audio.high-end
|
|||
|
|||
New psudo ABX ?
"bob" wrote in message
... On Mar 31, 8:20 pm, "Harry Lavo" wrote: 2) "If the person is good at it, he will probably reach a valid conclusion." This is a truism. No, it's actually a falsehood. No matter how good someone is at "open- ended evaluation of audio components" (whatever that is), we know that he will frequently come to demonstrably incorrect conclusions. That's what's wrong with "open-ended evaluation of audio components" (whatever that is): It produces a high incidence of false positives. But, as Steven noted, you are assuming that little problem away. I said "probably reach a valid conclusion". That implies more often than not. Where is your proof that a listener well-versed in live performances of acoustic instruments in a variety of venues (my qualifiers to the above statement) cannot be right more often than not, even in a sighted comparison. The possibility of bias does not mandate bias, and the possibility of error does not mandate error. This is a common objectivist oversight. |