On Thu, 12 May 2011 05:28:20 -0700, Arny Krueger wrote
(in article ):
"Audio Empire" wrote in message
ScottW seems to think that audio is somehow a "special
case" and is more than just a complex waveform.
Its a common audiophile myth - they also deny that Fourier was right about
complex waves being logical decomposable into collections of sine waves.
I have not been able to convince him: 1) that amplifier
performance is pretty well understood, 2) that total D +
N plots tell the whole story about distortion and noise
levels with respect to signal, and 3) aggegate D + N
levels of about -115 dB below the rated output or more
are, for all intents and purposes, inaudible.
I strongly disagree with the -115 dB number. It is way too tough.
There are two numbers that seem to relate to audibility:
-100 dB or "The -100 dB rule" which says that any artifact that is 100 dB or
more down is unconditionally inaudible.
I was purposely conservative with this to be "safe" because most modern,
solid-state amps (and some tube amps) "hover" around this figure or greater.
-80 dB, or "The -80 dB rule" which says that even when "only" -80 dB,
virtually all real-world aritfacts are inaudible.
Sounds reasonable.
Then there is also the -60 dB rule which represents the typical performance
of the best analog tape, LP playback and/or vacuum tube electronics, This is
often audible, but potentially tolerable and perhaps to some, euphonic.
Some distortion is even inaudible when analog tape is driven, momentarily, to
as much as +3 dB. However driving the tape to anything higher than that IS
VERY audible.
So -100 dB is unconditionally audibly perfect, -80 dB is generally more than
good enough, and -60 dB is often audible, but might not sound all that bad.
The actual FR artifact (linear distortion) detection ability of the human
ear was provided by Clark in his ground-breaking (peer-reviewed) ABX
article, and is online at http://home.provide.net/~djcarlst/abx_crit.gif .
Yes, as you say, it turns out that the human ear is quite insensitive to that
kind of amplifier distortion. A French tube amp (don't remember the name
exactly. It was something akin to Joule, IIRC) that was all the rage a few
years ago and was hailed by the audiophile press on both sides of the Pond as
being the world's best amplifier. When somebody measured it, it was found to
have more than 2% THD at 10 Watts and more than 10% just short of clipping!
The amp clipped at around 70 Watts as I recall. The publication even
requested a second sample (thinking the first must be defective). Nope, the
second one measured exactly the same.
While these seemingly high levels of distortion (-60 to -80 dB) may not be
audible as distortion per se, I do wonder if they might not be audible in
DBTs between amps in the -60 dB to -80 dB and amps that measure at -100 dB
and better? I'm merely speculating here, as I have no idea.
This chart shows the size and center frequency of FR artifacts that are
unconditionally inaudible. Typical JND's are about 3 times what it shows.
The chart has a built-in safety factor of 3:1 or so.
The ear is much more tolerant of linear distortion than nonlinear
distortion. The unconditional inaudible FR variation of 0.1 dB amounts to
being about 1% distortion. Typical variations on the order of 1 dB in the
midrange are audible, which amounts to 10% distortion.
Yep.
For comparison there are essentually no speakers that are free of audible FR
artifacts, and essentually none that have all nonlinear distortion better
than 40 dB down over a reasonble range of frequencies and SPLs. Under ideal
conditions, some speakers have most artifacts down close to -60 dB over a
fairly limited and incomplete range of frequencies.
There is a belief of some audiophiles that speaker distortion is somehow
more euphonic than the distortion in electronics, and this is generally not
true. Analog tape probably has a combination of nonlinear and linear
distortion that is the closest to euphonic of all common audio situations.
I certainly have enough experience in that realm to confirm that. I recorded
for years using a pair of Otari MX-5050s half track, 15 ips recorders.