Of course, the person at the controls is not passing the test at the same time... And it's blinded, as really blinded: blindfold goggle.
For the sequences, it was generated as randomly as possible: lines of ABABABBBBABAAABABAAAABBBABABABABBBB then cut anywhere in-between and put on the participant's sheet. I'm not tempted to use a computer to generate random sequences because a computer is not really able to generate true random stuff, last time i checked.
ABX testing method is widely used by multi-billions pharmaceutical companies for decades. In fact, it's a sina que non condition for FDA approval.
My methodology for this particular test may be flawed, but i don't think the ABX method is.
Or prove me wrong. 😉
The more we learn, the less and less this exercise resembles anything remotely like FDA clinical trials.
Odd that they were referenced to give some sort of credibility to the exercise.
Dppler9000, i think the more you read that thread, the less you understand the big picture. Maybe it's time to take a break and start a DIY project of some kind? Just avoid any DAC project, uncertain future.
;-)
;-)
You referenced FDA clinical trials on page 1 of your thread to provide credibility to your ABX approach.
Are they related or are you going back to the personal attack/deflection mode?
Are they related or are you going back to the personal attack/deflection mode?
There is no big picture.Dppler9000, i think the more you read that thread, the less you understand the big picture. Maybe it's time to take a break and start a DIY project of some kind? Just avoid any DAC project, uncertain future.
;-)
There is a group of guys in Montreal engaging in exercises that cannot be sensibly generalized.
It is a rather small picture.
You have enough samples now to extract statistics.
The AB test should have a binomial distribution, as you only have 2 outcomes, just like a coin toss. So the 95% confidence interval, assuming its a unbiased random pick, is 0.5+/-1.96*( 0.5 * 0.5 / 263)^0.5 or p= 0.5 +/-0.0604 or [0.4396 to 0.5604] which is what you got (actual 0.475).
Now for some semantics. If your null hypothesis was "there is a difference", then you are within 95% confidence of a random result, so you cannot statistically say "there is a difference". For most people the converse is also true that "there was no difference" detected.
Great experiment Jon, I can't believe how long this discussion has persisted. Witches, Spanish Inquisition, magic cables, golden ears, electron spin, untrained listeners, incorrect music, magic mushrooms, still no snacks, unquantifiable qualities ...
Start some more 🙂
I'm kind of afraid now to publicly discuss (at least on diyaudio.com) on that upcoming 4th test. I was virtually crucifed for something trivial as digital-to-analog converters...
But anyway, thank you DonVK for the good words, it's sincerely appreciated.
Come on, don't be put off by the big bullies, that's what they want 😉. It's evident that quite a few people are interestedI'm kind of afraid now to publicly discuss (at least on diyaudio.com) on that upcoming 4th test. I was virtually crucifed for something trivial as digital-to-analog converters
You referenced FDA clinical trials on page 1 of your thread to provide credibility to your ABX approach.
Are they related or are you going back to the personal attack/deflection mode?
Doppler, all along i was counting on the Earthworks M50 alone to provide credibility to my test. 😉 They're expensive calibrated mics!
Keep your smile, nobody died here. Only an illusion on Audiophilia.
I'm kind of afraid now to publicly discuss (at least on diyaudio.com) on that upcoming 4th test. I was virtually crucifed for something trivial as digital-to-analog converters...
It isn't the DACs, it is the "tests" themselves that are trivial.
You are too modest - you have done more than three of these - it is at lest four or five.
All with the same ex-ante assurances that differences will be heard.
All with the same guaranteed null confirmation approach.
All with the same level of "SHOCK" that the results indicated no audible differences.
if you are going to rely on these exercises to guide your purchase decisions, save the time and money, and go buy a rack system.
Come on, don't be put off by the big bullies, that's what they want 😉. It's evident that quite a few people are interested
You know what the 4th test is about, right, Scott ?
Well, let's hope there will be some positive identification this time, otherwise that will be an earthquake. 🙁
You are too modest - you have done more than three of these - it is at lest four or five.
More than three? Do i work while i'm sleeping? I didn't get the memo.
if you are going to rely on these exercises to guide your purchase decisions
Damn, doppler, you think all my efforts are motivated on a personnal basis ? For purchase decisions ?
It's R&D of commercial nature. It is from the past 10 years now.
I wouldn't put that kind of energy and money on a hobby, i can tell you that.
Come on, don't be put off by the big bullies, that's what they want 😉. It's evident that quite a few people are interested
I don't consider any of them as ''big bullies''... I just think it's a normal human reaction to be angry and/or in denial when someone is telling you that your past decisions were based on a lie, on an illusion.
The deeper you were linked to those past decisions and energy/money invested... The more it hurts.
Yep, not your average home listening room, I thought these were observant people 🙄
For the little story, back in 2009 i built from scratch 7,000 sq.ft. facilities that includes 4 rooms dedicated for audio test: the big one you see in the pictures, but also 3 clone rooms that were planned to be used for.... blind tests.
Ironically, the 3 smaller rooms were never used for blind tests and are now converted in office spaces. But still useable, if needed.
Doppler, all along i was counting on the Earthworks M50 alone to provide credibility to my test. 😉 They're expensive calibrated mics!
Keep your smile, nobody died here. Only an illusion on Audiophilia.
You don't need a calibrated mic to do level matching.
//
You don't need a calibrated mic to do level matching.
//
I know, that was intended as a joke. 🙂
The deeper you were linked to those past decisions and energy/money invested... The more it hurts.
It's amazing sometimes how strongly opinions formed with 70's-80's technology are held on to. Too much to lose in engaging in any exercise where the past beliefs might be questioned.
Were your tests 'calibrated'? Level matching? Trained listeners? Length/type of music samples? I ask because others may, as they have for the OP's test.
Calibrated: What type of calibration are you interested in?
Level Matched: I usually like to turn the level up and down during listening, and certainly at the request of or under the control of a listener. It helps to show that what we are listening for in the case of DACs is independent of level over plus or minus 4dB or more. We could do it level matched, of course, and level matching would be necessary if levels were to be fixed.
Trained Listeners: If they initially don't hear any difference, my usual protocol is to ask them to listen to the cymbals closely or something else closely and try to remember the details of what that sounds like just before we switch to the next DAC. I don't tell them what it sounds like, but I ask them to think about whether cymbals sound like bursts of hissing noise, like a real cymbal, if they can hear the sound of a drum stick hitting the cymbal on the attack, and things like that. Asking such questions and asking listeners to try to remember the details while DACs are being switched is the kind of training they receive initially. Later in the activities, its up to them to do it without any coaching.
Length/type of samples: For comparing DACs its not too critical. I would probably choose to use a style of music the listeners like to listen to, or at least that they don't actively dislike. Hi-Res source material with undistorted cymbals seems to make it easier for listeners, and I avoid material that is too distorted.
Last edited:
The AB test should have a binomial distribution,
But this is an ABX test, not AB.
dave
Training is the big factor - the problem is, of course, that there has to be someone who actually can identify a difference to clue people into it. If this is missing then we get the results that Jon seems to produce on every blind ABX test.Calibrated: What type of calibration are you interested in?
Level Matched: I usually like to turn the level up and down during listening, and certainly at the request of or under the control of a listener. It helps to show that what we are listening for in the case of DACs is independent of level over plus or minus 4dB or more. We could do it level matched, of course, and level matching would be necessary if levels were to be fixed.
Trained Listeners: If they initially don't hear any difference, my usual protocol is to ask them to listen to the cymbals closely or something else closely and try to remember the details of what they sound like just before we switch to the next DAC. I don't tell them want they sound like, but I ask them to think about whether cymbals sound like bursts of hissing noise, like a real cymbal, if they can hear the sound of a drum stick hitting the cymbal on the attack, and things like that. Asking such questions and asking listeners to try to remember the details while DACs are being switched is the kind of training they receive initially. Later in the activities, its up to them to do it without any coaching.
Length/type of samples: For comparing DACs its not too critical. I would probably choose to use a style of music the listeners like to listen to, or at least that they don't actively dislike. Hi-Res source material with undistorted cymbals seems to make it easier for listeners, and I avoid material that is too distorted.
But did I read that in his test some participants got non-null results for amplitude differences of 0.2dB and this was before they were told there was a measured difference?
- Home
- Source & Line
- Digital Line Level
- DAC blind test: NO audible difference whatsoever