Jan, are you coming to AES (Oct2-5), RMAF(Oct 10-12) AND Burning Amp (Oct 19)??? That would be a pretty sweet roadtrip.
edit: sorry OT. But thought Jan would join the debate.
edit: sorry OT. But thought Jan would join the debate.
SY said:
Nice argument. 🙂
The answer would be yes, but only if the level difference is not swamped out in the results by missing sensitivity of participants beeing not used to hear under double blind conditions.
I remember that Lipshitz was a strong advocat for having a RIAA-correction curve with better accuracy than 0.1dB to avoid sonic degradation.
On the other hand most scientific researchers i´m aware of, claim that you have to train your listeners not only in hearing differences in general, but especially also in hearing differences under double blind conditions to get useful results.
If they are trained they can reach an incredible level of sensitivity in DBTs.
If listeners are not trained in this regard there is strong evidence that even much larger differences would remain undetected (in a statistically sense).
SY said:
Double blind test, between eggs and tomatoes. 😀
"Close your eyes. Now we throwing A. Tomorrow we'll throw B. The next year please report the difference"...
Quite difficult; The thrower must not be able to feel the difference in shell texture or shape of object to prevent an unnatural bias of throwing one harder than the other.Wavebourn said:
Double blind test, between eggs and tomatoes. 😀
"Close your eyes. Now we throwing A. Tomorrow we'll throw B. The next year please report the difference"...
Worse, the supplier of the "ammunition" must not be allowed to see which kind has been supplied to prevent another form of bias.
And we haven't even discussed weight or whether either needs to be organic.
For a truly valuable answer the target must be convinced not to duck the on coming missiles.
Do they make white tomatoes? Or do we paint the eggs red?
Oh dear, this will be far more difficult than imagined. 😀 😀 😀
scott wurcer said:
(In response to me saying "... then how can I consistently get 10/10 in a string of A/B/X tests based on only a feeling, when I can't hear any difference between A and B no matter how hard I try?")
Testing was done using software presumably with level-matched samples so was as blind as I could hope for. It was more that I could "feel" the difference rather than hear it in any audible way, which I hastily concluded was because it was following a different pathway in my brain (used for location rather than perception of sounds). It's a known effect, no references for the audio case on hand, but for a good explanation of the visual equivalent look at:
http://en.wikipedia.org/wiki/Blindsight
That sort of thing makes a compelling case for our ears being able to sense things that are unavailable to our sense of hearing, and if that is true (which it seems to be) then it opens the door for a lot of this "subjectivist BS".
BTW, if my memory serves me correctly it was a winabx comparison between the original and 256kb/s MP3 of the well-known "castanets" sample. Perhaps those with golden ears can easily hear a difference, but for the life of me I couldn't, which is why I was so surprised by my first result (something like 9/10). Thinking it was luck I kept going, and after a while it was easy to consistently get 10/10 right, while growing more aware of how I was doing it. The original "felt" more "real" even though all 3 samples (A/B/X) sounded absolutely identical, and in the absence of any audible difference I was going with my best guess. Towards the end of the tests I had grown more confident about the reliability of this "feeling" but that never spilled over into any difference in sound. At lower bitrates (eg 128k) I could hear a difference.
As for level matching in general, I must admit I don't in most blindy subjective testing I do. It's not even possible for comparisons between speakers. The subjectivist take on this would probably be that tweaking the volume control is a valid part of the comparison, it kind of chooses its own level depending on the equipment - with some equipment needing more volume than others to depict the same level of detail...
I'm afraid you lost me. If an ABX test for sonic differences is properly designed only the ears (and possibly a kinesthetic sense of deep bass notes) are being used.adx said:
You can score 10 out of 10, but hear no difference? So you are proposing a new sense, an Extra Sensory Perception?
I am not trying to discredit your feeling, I simply don't understand what you are saying. The Wikepedia site is no help here as it proposes a phenomena generally not recognized in the scientific community. Please expand your post.
That invalidates the results.adx said:
So any subconscious perception / bias of what's better can be adjusted loudest until you get a high score. Sorry, that also makes it worthless as far as I'm concerned.adx said:
Well, maybe, maybe not. Consider this experiment: I have two boxes of gain which I encapsulate to prevent identification. They're coded A and B (maybe not the most original naming...). I send them to a test subject.
The subject is free to insert and remove them from his system at will, play whatever material he wants, diddle the volume control all he wants. Both the time that the boxes are used and the volume settings are automatically recorded. After a predetermined time (say, a month) the boxes are returned and the results recorded. The boxes are randomly recoded and the experiment is repeated, either with the same subject or another subject. After several go-rounds, we find that one of the boxes, no matter what the coding, is used more than the other. Or we find that one of the boxes always seems to be run at a higher volume setting.
Or maybe there's no difference at all.
Would you consider this a valid controlled test?
The subject is free to insert and remove them from his system at will, play whatever material he wants, diddle the volume control all he wants. Both the time that the boxes are used and the volume settings are automatically recorded. After a predetermined time (say, a month) the boxes are returned and the results recorded. The boxes are randomly recoded and the experiment is repeated, either with the same subject or another subject. After several go-rounds, we find that one of the boxes, no matter what the coding, is used more than the other. Or we find that one of the boxes always seems to be run at a higher volume setting.
Or maybe there's no difference at all.
Would you consider this a valid controlled test?
hermanv said:
The phenomenon of blindsight is a recognized, documented, well accepted and well explained phenomenon in the scientific community. Likewise, the ability to report audible perceptions without being conciously aware is.
It IS quite strange at first sight, and I know the feeling that it must be hocus pocus. But it isn't. There is a lot of processing going on in your brain of which you are totally unaware, but the result of which suddenly 'pop ups'. Very simple example: if you meet a stranger, you sometimes take an instant liking to that person, or an instant dislike. Why? Most people are unable to explain it. They do knot know which signals and inputs have been received, how they were processed and matched to earlier experiences and memories. The result, popping up, is 'dislike'. Or 'like'. Unexplainable to them, although of course there does exist a logical explanation.
I have read about tests with blindsight, where people scored 10/10 on identifying cards although they were blind. They said so, but the tester said: 'just guess'. 10/10 correct. When confronted with it, the subjects often became very angry because they were convinced they were cheated. After all, they were blind, for Pete's sake! But their vision input still is intact, the processing to 'decode' it is still intact, *only* the connection to the concious part is damaged. With dramatic results.
Since you cannot selectively damage people's brains for scientific research (not legally at least), the research is done with people that have a specific, preferable very local, brain damage, and only with full consent. By learning about the inner workings of the brain, previously unexplained phenomena fall into place.
Jan Didden
Brett said:
That automatically disqualifies blind listening tests between loudspeakers or any other audio DUT incapable of wide band matching to less than 0.1 dB. That or the statement isn't iron clad universal as you appear to believe.
Jan; I seem to remember you being an advocate of either double blind or the "if you can't measure it it doesn't exist" school.janneman said:
Your statement above makes all ABX tests meaningless. How can you support a methodology that has no meaning if the real reasons we like dislike a component or sound could be blindsight. Sure sounds subjective to me.
Your example of like or dislike of a person has indeed been well researched. Blindsight is not given as the explanation; Facial cues, sound tremors and body language neatly explain the phenomena.
You seem to want to argue both sides of the debate at the same time. Please remember that Wikipedia handy as it is, is not a scientific document anyone can post pretty much any opinion.
janneman said:
I think this relates to my comments about auditory savantism. In a more concrete context there are people that can multiply two ten digit numbers together and just recite the answer instantly. So yes, possibly there are perceptual things going on that we know little about.
OTOH I still think it is totally fair to say someone has the ability to "unconsciously" detect +- .1dB frequency response variations then after training this perception passes into the conscious. So my first contention still stands, if the known effects that can skew listening tests are not eliminated no conclusions can be drawn. To conclude out of hand that you could never train yourself so that your "feelings" became perceptions is a mistake IMHO.
isn't that what's clever about the Meyer-Moran test? They were trying to detect a difference. By proving no difference, then nothing else can be a factor either - the two paths are identical (perceptually). they must have been the same level (or whatever quality of sound we want to define.)
Still looking for the acoustic measurements on the various systems.
Still looking for the acoustic measurements on the various systems.
Iain McNeill said:
From a scientifically point of view it simply doesn´t work that way.
Of course theres is room to argue what part of an experiment could have been used as a control _after_ the outcome, but a carefully done test design must controls for these parts of the experiment.
So it must include a negative control to make sure, that a detected difference is based _only_ on a audible difference and it must include positive controls to proove that a sufficient hearing sensitivity level was reached by the participants under the specific test conditions.
If the controls were ommitted the outcome is totally indifferent as you can only speculate about the reasons.
Maybe in the M/M test there was no audible difference but maybe the reason was simply that the detection threshold was to high for the difference presented.
scott wurcer said:
I think what happens when you 'unconciously' detect the .1db difference in freq response, it wouldn't necessarily pass into conciousness in the sense that you would be aware that one amp has +.1db wrt the other. Based on what I've read, I would expect that you would select more or less consistently one amp over the other, but without really knowing it is because of the +.1dB.
I agree that training can help you to identify influences and reactions you were not aware of before.
Jan Didden
Some clarification on my blindsight post: I wasn't suggesting there was anything weird or even subconscious going on. I think some of the apparent weirdness stems from a widely held belief that "ears = hearing". Not everything your ears measure goes on to form the sense of hearing, conversely not everything that you hear comes from your ears. It is the same for all senses. The best example is dreaming - your senses are working overtime (as confirmed by brain measurements) in the absence of any input from the sensors.
I suspect blindsight wouldn't be accepted, or at least would still be put down to "something weird", if it weren't neatly explained by examinations of the structure of the brain which show that a portion of the signals from the optic nerves go to different places than are responsible for visual imagery (that's the TV documentary version). There are also a lot of connections from "higher centres" back into the visual system. Presumably there's good reasons for these, as simple as things like energy saving - caching of sensory decoding, eg once you know what something looks like you can fairly safely imagine it rather than see it (also the TV documentary version). The same appears to apply to hearing, but a lot more research effort goes into vision than hearing.
The implications of this for audio are obvious. It means that peoples' senses aren't "fooling them" when they say something sounds better because of (say) placebo effect - it actually does sound better, as reported by the sense of hearing, even if it is coming by a different pathway than from the ears. It also means that our ears can sense aspects of sound that our sense of hearing never gets hold of, which further messes up any tests based purely on hearing (as opposed to what our ears sense, which isn't the same thing). I'm not saying such tests are invalid, just that they are not the complete answer, even in a purely scientific world.
Getting back to SACD vs CD, I think it boils down to whether ultrasonic signals are audible (if you get what I mean). I would be happy to ignore the extra bits for the mean time, because all they do is define the noise floor - check out MegaBitMax 'dither' - it is amazing what can be done with 16 bits these days, and the old-school 16 bit noise floor (tpdf or gaussian dither) is amazingly quiet anyway. But there is a lot of anecdotal evidence for audibility of 20kHz frequencies in complex sounds, but I've never seen any real proof, and the M&M result doesn't do anything to help with that. I was voicing some speakers a couple of days ago, they had a nasty 17.5kHz peak which I decided I needed to tame, but I couldn't hear a difference at all (perhaps an ABX test was in order). Disappointing but unsurprising in hindsight, it has been a while since the 19kHz FM stereo subcarrier has annoyed me.
Of course SACD is better than CD because it defines a lower end of acceptability that picks up where CDs meet their limits, for the entire signal chain really. All this attention to 100dB+ dynamic range and >>20kHz bandwith is WAY more than most CDs get, and means such nasties as mixing in 16 bit and truncating are completely unacceptable, even if the end result happens to work just as well on CD as SACD. That's pretty much M&M's conclusion.
I suspect blindsight wouldn't be accepted, or at least would still be put down to "something weird", if it weren't neatly explained by examinations of the structure of the brain which show that a portion of the signals from the optic nerves go to different places than are responsible for visual imagery (that's the TV documentary version). There are also a lot of connections from "higher centres" back into the visual system. Presumably there's good reasons for these, as simple as things like energy saving - caching of sensory decoding, eg once you know what something looks like you can fairly safely imagine it rather than see it (also the TV documentary version). The same appears to apply to hearing, but a lot more research effort goes into vision than hearing.
The implications of this for audio are obvious. It means that peoples' senses aren't "fooling them" when they say something sounds better because of (say) placebo effect - it actually does sound better, as reported by the sense of hearing, even if it is coming by a different pathway than from the ears. It also means that our ears can sense aspects of sound that our sense of hearing never gets hold of, which further messes up any tests based purely on hearing (as opposed to what our ears sense, which isn't the same thing). I'm not saying such tests are invalid, just that they are not the complete answer, even in a purely scientific world.
Getting back to SACD vs CD, I think it boils down to whether ultrasonic signals are audible (if you get what I mean). I would be happy to ignore the extra bits for the mean time, because all they do is define the noise floor - check out MegaBitMax 'dither' - it is amazing what can be done with 16 bits these days, and the old-school 16 bit noise floor (tpdf or gaussian dither) is amazingly quiet anyway. But there is a lot of anecdotal evidence for audibility of 20kHz frequencies in complex sounds, but I've never seen any real proof, and the M&M result doesn't do anything to help with that. I was voicing some speakers a couple of days ago, they had a nasty 17.5kHz peak which I decided I needed to tame, but I couldn't hear a difference at all (perhaps an ABX test was in order). Disappointing but unsurprising in hindsight, it has been a while since the 19kHz FM stereo subcarrier has annoyed me.
Of course SACD is better than CD because it defines a lower end of acceptability that picks up where CDs meet their limits, for the entire signal chain really. All this attention to 100dB+ dynamic range and >>20kHz bandwith is WAY more than most CDs get, and means such nasties as mixing in 16 bit and truncating are completely unacceptable, even if the end result happens to work just as well on CD as SACD. That's pretty much M&M's conclusion.
adx said:
No digital console I know uses 16 bit internal processing. On last look many years ago my recollection is double that as the norm for exactly the reasons you imply. Again, one of M&M's major methodological flaws was failing to confirm the budget players - 2 of the 3 units used - provided a valid, full rez SACD output.
The Sensible Sound's technical analysis of the Yamaha here: http://www.audiotrends.com.au/pdf/Yamaha/Reviews/Yamaha_DVD_S1500_SS_Review.pdf . Read it yourselves and confirm my read, however paraphrasing the author; 'little better than 17 bit performance, falling back to 16.5 bit at full scale with slightly less than 16 bit distortion performance.' If correct that one finding to me poisons all the compiled data. The author was also incapable of determining if the player internally converted from DSD to PCM, "in which case any advantage of SACD signals having no digital signal processing is rendered moot."
Re: the Pioneer, some report it converts to PCM: http://www.goodsound.com/equipment/pioneer_dv563a.htm. To quote "The difference between the two players could arguably be attributable to the fact that while the Sony processes the native DSD signal, the Pioneer converts DSD to PCM before processing. Pioneer does this to save the cost of making a player that includes two processors." A full technical analysis here by Charles Hansen: http://www.audioxpress.com/reviews/media/Hansen1104.pdf
My read is M&M's work is a comparison of three specific players against the HBB CD recorder's ADA circuit with no wider applicability to SACD vs. CD and a textbook example of the effect on experimental protocol of 'knowing' the results before collecting data.
One of the basic precepts in science is paraphrased as follows: "I can teach you to perform a test, once taught, you will get the same answer as I did every time."
I am of the subjectavist school, I hear things that matter to me, slight differences in cables or capacitors for example. I am happy enough that all of what I believe is not fully covered by scientific application and consistently tell people to form their own opinion. It's only me that needs to be pleased with the end result of my tests, I feel no need for any apologies.
However my professional training does insist on my first paragraph above, can someone teach me "blindsight" or show me an example of where I'm already using it?
My subjectavist analysis of how things sound is a repeatable (to me) exercise using hearing, I can with some difficulty describe what I hear. This difficulty is not due to any ephemeral nature of what I hear, but mostly due to a gap in language. No such thing as electronic reproduction shortcomings existed as English was being developed over the centuries.
I am of the subjectavist school, I hear things that matter to me, slight differences in cables or capacitors for example. I am happy enough that all of what I believe is not fully covered by scientific application and consistently tell people to form their own opinion. It's only me that needs to be pleased with the end result of my tests, I feel no need for any apologies.
However my professional training does insist on my first paragraph above, can someone teach me "blindsight" or show me an example of where I'm already using it?
My subjectavist analysis of how things sound is a repeatable (to me) exercise using hearing, I can with some difficulty describe what I hear. This difficulty is not due to any ephemeral nature of what I hear, but mostly due to a gap in language. No such thing as electronic reproduction shortcomings existed as English was being developed over the centuries.
- Status
- Not open for further replies.