Testeur ABX

[Ragnarsson]

A propos de test en double aveugle, je suis tombé, en cherchant tout autre chose, sur un billet de Michael Fremer (Stereophile) qui évoque un de ces tests, réalisé à l'AES:

"You would like to see double blind A/B/X testing because you are SURE of the results.

You know the conclusion. You are not really interested in the experiment. That's why when this little experiment, which doesn't meet your standards, seemed to disprove your cherished beliefs, rather than saying maybe there's something here, you chose to deny its validity and to suggest a different test. Maybe we will do just that in the future.

And were I to take and pass your test, I am certain you'd do the same thing again. You'd move the goalposts. Or if a large group took the test and only I got it right, my result would be tossed and I'd be declared a "lucky coin," which is what happened in an amplifier double blind test I took at an AES.

The test was set up specifically for me to fail because someone challenged ME to a double blind test. Someone who claimed that all amplifiers that measure the same sound the same.

Five amplfiiers were chosen (not by me) and used for a test set up by someone else: the person who challenged me to take his test.

I got 5 of 5 identifications correct. Stereophile editor John Atkinson got 4 of five correct. But among the many members of the AES who took the test the results were statistically insignificant. Did that prove the contention that all amplifiers that measure the same sound the same?

Well, actually not because for some reason the test producer chose to include a Crown DC-300 known as an "ear-bleeder" and a VTL 300 a warm, vacuum tube amplifier. These amplifiers sound and measure very differently yet the AES members, most of whom were audio engineers could not distinguish between them!

Now if the test creator's point was to show that I'm a bad listener his test backfired, since I was easily able to hear those two.

Though I passed the test I got no credit and my detractors said "well there weren't enough samples. It wasn't a valid test. You don't understand statistics, etc." Of course I didn't create the test! And had I failed it do you think those same people would have said "There weren't enough samples, it wasn't statistically valid?" Of course not! They'd have said "You failed. All amplifiers do sound the same." That's how it works."

Peux tu s’il te plait donner le lien d’origine afin de pouvoir trouver l’envers du décor, le point de vue de l’autre partie, afin de ne pas introduire un biais dans l’interprétation et les conclusions.

Un test valide doit se faire avec un minimum de participants (15 tous ayant eu leur audition testée comme normale), sur un nombre suffisant d’échantillons musicaux (20), le tout pour chaque paire à comparer (à SPL totalement identique donc gains ajustés), sur des enceintes aux caractéristiques spécifiques dans un local fait pour (dimensions et RT).
Sinon le test est ni reproductible ni valide.

La mise au point des protocoles d’évaluation de qualité subjective a fait l’objet de nombreuses études qui ont permis de les valider.

Deux amplis aux caractéristiques identiques sur signal sinusoïdal et charge ohmique pure, en régime établi sonneront pareil dans cette charge résistive au bruit près.