Bob Marcus wrote:

A perennial claim holds that much of what objectivists think they know
about audio is wrong because they rely on short-term switching

snip

Experiment #1 was conducted by David Clark and Lawrence Greenhill in the
late 1980s. Clark rigged up two black boxes. One was a straight-wire
pass-through. The other added 2.5% harmonic distortion to the signal.
M
snip
Results were null. In a quick-switching ABX test,
however, subjects were able to tell the difference between a clean
signal and one with 2% distortion added.

Experiment #2 was conducted by Tom Nousaine in 1996. He prepared two
sets of CD-Rs. One set of CD-Rs was a bit-for-bit copy of a commercially
released song. The second set added 4% harmonic distortion to the song.
snip
Using a looped 6-second extract of the song, this subject was
able to score perfectly.



The above apocryphal stories are hardly dispositive of this issue.

The primary factor is "added distortion." Dr. Earl Geddes and others
have shown (in JAES peer reviewed articles, btw) that the mere addition
of or level of "distortion" does not correlate to the perception of the
existance of said distortion.

In the case of the Clark/Greenhill "test" it is unclear *as well* if the
above mentioned issue was at play, and if the gear of the day that was
actually used was capable of distinguishing the intended difference, no
matter how many audiophiles were involved. How the added distortion was
produced is another issue as well...

In the case of the erstwhile Noisaine's work, some years back he sent me
a CDR of some "musical" work with various levels of 'unknown changes to
the signal' (presumably some sort of distortion) which imho due to the
nature of the musical segment itself was spectacularly poor for
discerning *any* differences. Yes, there are recordings like that - ones
that seem to sound "good" no matter what the heck they're played on or
with. I have some.

By way of example and illumination of the last point - when one is
"tweaking" the values of a passive xover on a speaker system, changes of
small values of C or R tend to be measureable, but completely inaudible
by ear when listening to music. Put on pink noise, and suddenly very
small changes of C and R are fairly easy to hear! The point is that some
'signals' be they music or test signals are appropriate to discern
certain differences or changes, while others are not useful at all.

BTW, In practice, when the xover in the above example is set optimally,
the perceived impression of the speaker system as a whole is usually
improved over the less optimal alignment.

In summary, short switching is good for making some determinations. Long
term listening is good for making other determinations. Imho, neither is
perfect or sufficient alone. All methods involving human perception are
to some extent variable, and less than definitive - except to the extent
that in a general sense it is possible to determine many things, (for
example its fairly definitive that probably no one can hear 100kHz...)
using assorted 'tests', but as of yet, not all things.

But, why does this issue matter to anyone??

_-_-bear