[OFF-TOPIC] From: How do you listen to an ABX test?

2015-04-10 06:56:58

i find people strange .
you trust your ears when you choose speakers . Yes  we all look at specs  and  all the numbers . but at the end of the day it boils down to what sounds best to us .  in our environment in our homes .
Same thing with amplifiers  dacs ?
The whole journey begins with what we hear ,or heard in a demo room somewhere.
The rich walk into a demo room like how it sounds so much they buy the whole setup.
The  savy  spend sleepless nights building speakers , amplifiers . a few gentlemen even have get their own dacs built  :point:
the passionate drive around listen to different systems
.
So WHY do we argue  that our ears are unreliable devices .  Yet the whole journey begins with what we heard .

[OFF-TOPIC] From: How do you listen to an ABX test?

Reply #1 – 2015-04-10 10:22:58

I agree that the average person and even just about any person won't hear a difference between amplifiers blindfolded, once you choose amplifiers of similar output impedance and REAL output, not advertised output. Because that should in all likelihood sort out the frequency response issues in MOST cases when you feed a speaker with sound. Hell a lot of people won't hear the difference between a competent and incompetent amplifier, hence the proliferation of incompetent 2000W amplifiers.

So my first issue is this: Is this carefully controlled situation the case in the real world? not at all. There are many cases where the same class AB BJT 50w amps have significantly different sized transformers and heatsinks. These are affected differently by speaker impedance curves, when played at higher levels and the music starts sounding life like, at lower levels probably not. I don't believe you have issue with this, you design amplifiers with large enough PSU and heatsinking, afterall... but when double blind tests control for all of the real world things above, they are not very realistic at all, or shall we say they apply for only those conditions.
Especially when they then state that:
THOU CANST NOT HEAR A DIFFERENCE IN AMPLIFIERS......... and in small text footer state...provided they are level and frequency response matched and don't distort... Is the science wrong in this case, No... the results in a context are just preached out of context.

My second issue is this: I do believe that statistics cannot accurately judge the significance of performance by small sections of the population. And I do not agree that people who apply statistics is above reproach, I also do not believe that statisticians are above reproach, because they often do not understand the underlying assumptions or trends in the systems for which they are asked to analyse data for. At varsity we had a top-notch stats department, yet whenever our biological data became too complex, or skewed, because there was a slant in a population or many interrelated populations, they were of little help, though not for a lack of trying. Typically in vegetation data across populations in space or time, we first have to objectively and/or subjectively classify data into distinct groups so that we may analyse the data accurately, be this with mathematics through ordination or regression analyses or statistics, typically we didn't do much stats, basically because there is no use. (This is an ongoing field in biological data analysis, there are people who still dedicate their life's work designing math, and programs to support that math, that take out the distortions the "old" math imparts on the data for natural systems.) We used statistics to assess one performance variable in a very closely controlled environment - which is why one would assume that for testing if humans could hear the difference in amplifiers, stats would be the way to go, I believe we are missing something.

Now if you take a complex system like hearing and perception and you want to run an experiment, you draw people from a wide variety of backgrounds - because people are much travelled and have different value systems. Even though typically they care perhaps enough to spend on music. This means you are violating the fundamental principle of probability theory. Because you cannot assume that all people have the same chance of getting the answer right or wrong - if you do... well then you assume that if the audiophile reviewer can get it right, everyone can. (i.e. if that Husain Bolt can run 9.9sec 100m then so can I - which if you look at me, even in the prime of my athletic ability, was patently ludicrous, but if you look at him and several of his buddies and the others who share the track with him...)

As evidence that this is correct, I point out that in Richard Clarke's extremely controlled amplifier experiments at the AES, Michael Fremer actually scored a perfect score and so did (if i remember correctly) John Atkinson, and in John Atkinson's amplifier AB difference test, several people who spent their working days evaluating audio equipment got perfect or near perfect scores. This correlation proves that the experiment violates the fundamental principles of statistics, there is too much variation in the data / sample of the population for a meaningful result to be obtained. Meaning the experiment was set up wrongly. I do not know if the studies you refer to, correct for this. I believe not, because when the data looks so tidy and normal, as it does when there is too much variation to properly analyse, nobody bats an eyelid.

[OFF-TOPIC] From: How do you listen to an ABX test?

Reply #2 – 2015-04-10 12:48:25

What about Fremer acing DBTs on amplifiers? Is his hearing golden-ear level or did it amount to luck?