Re: setting up ABX tests at home

bear <bearlabs@[EMAIL PROTECTED]
> wrote:
> Sonnova wrote:
> > On Mon, 11 Feb 2008 16:17:08 -0800, bear wrote
> > (in article <foqoi402tkd@[EMAIL PROTECTED]
>):
> > 
> >> Arny Krueger wrote:
> >>> "Greg Wormald" <greg.wormald@[EMAIL PROTECTED]
> wrote in message
> >>> news:folrst02p34@[EMAIL PROTECTED]
> >>>
> >>>> The very set-up of much ABX testing (which is most often
> >>>> short segments of music quickly switched), cannot test
> >>>> for what most music lovers are seeking--long term
> >>>> enjoyment of music listening.
> >>> This is a false claim that I've heard on and off for about 30 years,
now.
> >> Well Arny, where is the correlation between ABX tests and subjective 
> >>

> <snip>

> > onnects are so bad that they cause non-trivial frequency response 
> >>> variations, they are by definition, defective. 
> >> This would appear to be a true statement.
> >> But it is not.
> >> The reason that it is not true is that the "ABX" tests that have been
touted 
> >> as 
> >> providing this information are fraught with unknown variables.
> >> Let me list a few for the readers:
> >> - frequency response of the speakers
> > 
> > if you are using the SAME speakers for each component under test, the 
> > frequency response of the speakers is a constant, NOT a variable and 
> > therefore is irrelevant. If you are using different speakers for the
ABX 
> > test, then you're doing it wrong by introducing more than one variable
into 
> > the test.

> You are simply not comprehending the significance of masking due to 
> distortion(s) nor the effect of cascaded distortions.

> Put it another way, IF the distortions present are of a particular type
or class 
> then you can put anything you like into the signal chain of lower or nil

> distortion and most likely not be able to identify any change. Since you
have no 
> way to determine the significance or type/class of distortions in your
system, 
> you don't know anything about the results of an ABX test other than the
results 
> being SPECIFIC to the test you performed.

You're getting way too analytical.  It's admirable from a scientific POV 
to want to exhausively determine, a priori, whether a difference between 
two treatments is audible.  This is how mp3 codecs have evolved, for 
example.

In the common run of 'audiophile' culture, however, all that's required is

to test its many, many claims of ALREADY HEARD difference.  Which is to 
say, to repeat the anecdotal 'sighted' experiments, where differences are 
sometimes re****ted to be extravagantly large, this time using DBT 
conditions.  THis tells us whether the ORIGINAL claim was likely to be a 
re****t of a real difference, or not.