[Audio Empire]

On Tue, 10 May 2011 12:19:07 -0700, Arny Krueger wrote
(in article ):

"Scott" wrote in message


Sounds very reasonable. Now the question is how does this
apply to real world amplifiers. When we actually put the
vast world of amplifiers that have been commercially
available since the earliest claims of amplifier
transparency to the current times how many of them
actually have a measurable aggregate D+N levels at or
below a -115 dB below the rated output?

You're not following the discusison, Scott.

Please review the thread(s) and determine the correct number before
proceeding.




Anyone wanting to read a good paper on amplifier distortion and measurement
techniques, I recommend this one:

http://tinyurl.com/3dxqnm3

The maths are not so difficult that a layman can't follow the gist of the
article without understanding it.

BTW, Arny, to be fair to Scott. I'm the one who used the "around" -115 dB
figure as an example of inaudibility and that's an educated guess on my part
(and I only used it as an example). Most tests seem to express distortion in
percent, not dB, And I have not found the threshold of hearing listed
anywhere as a function of amplifier output. I know, for instance that 0 dB is
considered the threshold of hearing based on the Phon scale, and likewise 130
dB is the threshold of pain. If you invert that so that 0 dB represents the
threshold of pain then -130 dB equals the threshold of hearing. But I have
been unable to locate exactly how these numbers correlate to an amplifier's
output. Looking at the results from a number of amp tests from magazines who
use computer-based waveform analysis, shows most modern SS amps to be better
than -110 dB, so I picked "around" -115 dB to use as an example of
inaudibility. If you know the actual threshold of hearing for noise and
distortion in amps, I'd love to see it.

[Scott[_6_]]

On May 10, 12:19=A0pm, "Arny Krueger" wrote:
"Scott" wrote in message



Sounds very reasonable. Now the question is how does this
apply to real world amplifiers. When we actually put the
vast world of amplifiers that have been commercially
available since the earliest claims of amplifier
transparency to the current times how many of them
actually have a measurable aggregate D+N levels at or
below a -115 dB below the rated output?

You're not following the discusison, Scott.

Again with making it about me and not about audio.


Please review the thread(s) and determine the correct number before
proceeding.

If you don't like those numbers then you might want to take it up
with the person who posted them. The number I provided was..... oops
you snipped it. Why did you do that Arny? Why claim I am providing
eroneous information and then snip that information? Here it is again.
I'm going to take the liberty of quoting myself if the mods will allow
it. "Or more significantly below the possibility of audibility
whatever that threshold may actually be when talking about the
aggregate D+N levels and all the possible artifacts they may
represent."
So what number is that Arny? Here is a hint, no matter what number it
may be it is the correct number by definition.

[Audio Empire]

On Tue, 10 May 2011 17:41:36 -0700, Scott wrote
(in article ):

On May 9, 4:03=A0pm, ScottW wrote:
On May 8, 4:35=A0pm, Audio Empire wrote:



If all frequencies are represented equally, how would this NOT satisfy =
your
rather naive (the naivete comes from your belief that it is necessary t=
o
uncover and quantify all possible causes of harmonic and intermodulatio=
n
distortion) premise?

=A0I'm kind of shocked that someone would propose it's not necessary to
"uncover and quantify all possible causes" of distortion before they
can use the data of a finite set of incomplete measurements to declare
transparency.

I think we are getting off path here. The issue is not "all possible
causes of distortion" It's a question uncovering all actual audible
distortions. We don't need to search out some near infinite relm of
possible distortions. We can measure the actual distortions when an
amp is fed a musical signal and drives a real world speaker load. And
really, let's not forget that this whole discussion was spawned by my
challege of the assertion of scientific support for the belief in
amplifier transparency. If one wants to make *that* assertion then one
obviously needs the actual scientific support. One can't derive a
proof for such a belief in amplifier transparency based on a
corolation between measured amplifier performance nad human thresholds
of hearing as some have tried to do here. One can noty reasonably use
that approach as scientific support without first establishing that
the measurements represent a comprehensive set of meausrements of
distortions that may or may not be audible. If not then how does one
know that the meausrements they have represent a complete picture of
the amp's audible performance?


But that support does exist. Pick up any magazine that does wave analysis
tests of equipment and look at the D + N plot.

[Sebastian Kaliszewski]

Scott wrote:
On May 10, 5:42 am, Sebastian Kaliszewski
wrote:
[...]Scott wrote:

On May 6, 10:13 am, Sebastian Kaliszewski wrote:
Scott wrote:

On Apr 26, 6:19 am, Sebastian Kaliszewski wrote:
Scott wrote:

On Apr 23, 8:21 pm, KH wrote:
On 4/23/2011 9:07 AM, Scott wrote:
On Apr 22, 8:32 pm, wrote:
But guess what? That ain't the
case with amps. Here is a challenge. Set up any two amps of different
design, set them up for a null tests using an actual real world
speaker load. Then measure the respidule signal. If any of them are
-100db then I will concede straight up that those two amps are audibly
identical with that speaker. *IF* you can't do that you cant claim
both amps are -100db in total distortion with a real world speaker
load.
You do understand that "-100db" was an *EXAMPLE*, not a CLAIM? Talk
about posturing.
Fine then make a claim and support it.
It has been done. But specially for you:
the estabilished requirements a
frequency response:
+-0.2dB from 20Hz to 10kHz
+-0.5dB from 16 to 20Hz and from 10 do 20kHz
phase response:
+-5degree (most literature claims +-30degree, but some claim greater
sensitivity for some artificial signals; for real music it's +-30degree but
let's be generous)
nonlinear distortion:
HD: 2nd harmonic 2% (-34dB)
all the others 0.2% (-54dB) (most literature claims 0.3 to 0.5dB, but
let's
be generous)
IM: 0.1% (-60dB)
uncorrelated spurious tones: 0.01% (-80dB) (for A,B class amplifiers it
applies
only to power supply pulsation, for D class as well as DACs/ADCs there are
other
possibilities)
uncorrelated aperiodic (atonal) noise level dependent: 0.2% (-54dB)
noise floor (i.e. noise uncorrelated to actual signal): -80dBW
SNR: 90dB
OK this is a nice move forward. So these are the "established"
requirements for amplifier transparency. How were they established? Do
you have any references for this?
See those yoy ignored.

Nonsense! I did not ignore them. That is a completely unreasonable
thing to assert in a post responding to my acknowledgement of them.
I_asked_you_to_varify_the_assertion_by _citing_ANY_references_for_it.
So asking for varification is synonymous with ignoring something?
Really? Looks like a lot of obfusecation in place of substance.

You declared them irrelevant without looking at them except one.

Simply not true. Do you have a quote in context where I did this?

Short memory, Scott?

Who wrote: "Nice stack of irrelevant material only some of which would be
considered peer reviewed science."?


No
you do not because it never happened.

Except it did

This is why I ask that people
actually quote me in context while trying to represent my position. I
asked for references nothing more nothing less and you have provided
none to this point.

....



See pre WWII results.

show em to me. If you've seen em show em to me!

Those Fletcher-Munson curves are how old?

Once again you fail to support your alleged scientific evidence with
actual citations.

This one seemed to be obvious to anyone...

Fletcher, H. and Munson, W.A. Loudness, its definition, measurement and
calculation. Journal of the Acoustic Society of America 5, 82-108 (1933)




See recent 16bit/44.1KHz ADC-DAC
insert in the chain (which has obvious hard limits, all less stringent than
-100dB) -- it was null result published in peer reviewed journal.

What on earth does this have to do witrh amplifier transparency? I may
as well look at the latest Dow Jones Industrials!! This obfuscation at
it's weakest.

It demonstrates that complete destruction of signal below about -94dB is
undetectable by hearing. ADC-DAC chain simply destroys all information below
that level.

Sorry but you will ahve to show that the potential for distortion
artifacts arew the same for ampilifiers as AD/DA converters. So this
particular item is irrelevant. If you want to establish the thresholds
of human hearing you will have to cite the actual peer reviewed
studies for that. So far you have come up empty on that one.


You can postulate all manner of interactions of low level distortions,
but unless the sum or conjugate of all those interactions reaches above
the established threshold of audibility, then any amplifier with such
measurements will be audibly transparent.
OK then show me that is the actual real world case. Simple null tests
with all the real world amps using a real world speaker load will do
the trick. Oh waite a minute......Bob Carver already did that.
http://www.stereophile.com/content/c...allenge-page-2
"Bob's test hookup would show much more than frequency response
differences. In fact, one of his most interesting statements, for
those of the "every amplifier is the same except for frequency
response" school, was that varying frequency response between the 1.0
and the reference amp made up only about 25% of the significant
differences. Relative phase shift, source impedances (damping factors)—
in short, every electrical difference between the amplifiers—would
produce a signal at that test point between the Plus output terminals.
When the amplifier outputs were identical, in all respects, there
would be total cancellation—a null—of the difference signal. Bob's
goal was a 70dB null, or an 0.03% difference between the two amps.
And he was able to measure all of this, no? Where was the "mystery"
distortion you postulate?
Yes he was able to measure differences between all the amps in
question. Indeed there was no "mystery" distortion. So clearly I am
not making such a claim. That is just another in a long string of
misrepresentations of my position. What we have here are measured
differences between amps that clearly do not fall below the
scientifically established thresholds of human hearing.
The do.
Really? Do you know what amps I was talking about and how they
measured against each other?
Yes. You're talking about Carver Challenge as described
onhttp://www.stereophile.com/content/carver-challenge-page-2

They were no where near a null.
The were. At least that's what could be read from what they wrote in
Stereophile. Differences between amps at -70dB fit into that range. (60dB is
required in fact) unless one of those has PS noise above -80dB.

I think you may have missed the bigger picture here.

I did not.

The amps were
around -70db *after* carver spent two days tweaking his prototype.
*Before* he did any tweaking they were nowhere near that.

So?

If you don't understand why this matters then you did in fact miss the
bigger picture here. That being the world of actual production
amplifiers and their actual measurable differences in distortion. Oh
well.

So what? That's exactly your misrepresentation (now clearly stated) of mine or
others position... could there be amplifiers which are not transparent? Of
course, there certainly could be, gosh, some SET stuff has distortions in the
range of 10%.



The whole point of the exercise was to make one amp a sonic fascimile of the
other. Amp being copied is not claimed to be transparent.

If you are backing away form the broad belief in amplifier
transparency then we are done are we not and you have conceded the
point have you not?

Nope, I do not. But here we're discussion Carver challenge *you* introduced into
this thread. Carver challenge demonstrates just that it's possible to make two
completely different amps sound the same.

Otherwise, Carver challenge has little to do to amp transparency, but it was
neither me, nor anyone else but *you* who introduced that. You introduced that
to create a strawman, to sidetrack the discussion or what?

We are talking about three different amplifiers

Really? Carver challenge was test of two amplifiers:
Conrad-Johnson Premier Four & Carvers prototype one.
Now you introduce also Carver's final product. But OK.

all three of which were commercial products, two of which were SS.
These are real world amps. Clearly they all were quite different in
their measured distortion.

Err, what? At least two of them were similar down to -70dB.

So much so that one could not look at their
differences and make any reasonable claims of transparency for any of
them bases soley on those measured differences.

Err, where was the claim of transparency?

Strawman, again.



That anyone claims thal all amps are transparent is just your misrepresentation
of what others wrote.


Maybe you should go back to the begining of this thread and actually
read what it was I have been arguing against.

See above few paragprahs of yours.


AND
interestingly enough the new production unit based on the tweaked
prototype in fact achieved nowhere near the -70db mark that the
prototype achieved. So with the actual real world amps that you and I
could have bought there were no such nulls. Even the _70db doesn't
tell us anything without the listening tests that were done after the
fact.

-70dB tells us that the sum of *irrelevant* as well as *relevant* differences
was at -70dB. Thus relevant differences could not be above that -70dB.

Yes for those two amps. But even then we can't use that measurment as
proof of same sound between the amps because we can't determine what
was audible and what was inaudible.

But we know that what relevant was below 70dB.

That was my point. You don't have
suypport for any assertions of transparency from this particular test.


[snip metadicussion]
It was you who bring Carver challenge and it's you who shot it down as
unscientific. All of that simply unrelated to the question wether science is
supporting the belief in existence of amplifier transparency (that there are
transparent amplifiers out there, and are in fact non negligible fraction of
audio amp models "in the wild"). All I said is that science is supporting that
belief.

Well at least I brought *soemthing* to the conversation in the way of
actual tests. Not sure how this is a problem.

As did others. And in fact those others presented real hard numbers not talk
about such numbers. As you replied to some (while ignoring others), You can't
claim you didn't see them.

The realk problem which
you have not addressed is the missing scientific support for the
belief in amplifier transparency. so far you haven't come close. The
rest is merely more obfusecation. Thus we have these painfully long
posts.

Nope

[snip more of metadiscussion]
First you bring Carver Challenge and then dispute its applicability to matters
discussed.

No. I simply pointed out that it was the onloy actual data on any
amplifiers presented so far and as such does not support the assertion
of transparency.

False premise.

Not a premise. It's simple fact

False.


It is not the only actual data. Many more relevant data was presented in this
very thread.

I didn't see it. Citation please

You did, as you even replied to some of those (for example those presentned by
Arny).



We need account for *only* the distortions that have the ability to show
up as variability in the sound waves exiting the speakers,
at or above
the threshold of hearing.
And how do you know what those distortions are in total?
Well gee, you spent a good deal of time discussing how Bob Carver did
exactly that. Doesn't seem that difficult.
Fair point. But by his measure very few if any amps are actually
transparent.
But his measure is overly sensitive.
Of course because it was said so in Stereophile.
Not. It was because actual audibility tresholds of many distortions are in fact
much less stringent (some, like for example time delay are 70dB easier).

But since you haven't shown us any support for the thresholds you
allege *and* you have failed to make a proper corolation between the
null in question (much less the nulls between all the actual
production units involved which were nowhere near -70db) and the
thresholds you asserted, you have not made any kind of a case so far.

Sorry, but I assumed that it's simple for everyone participating in the
discussion. Since it's all very basic inference.

This is a weak excuse for failure to support your assertions of the
thresholds of human hearing with actual scientific literature. The old
"obvious" card. VERY unscientific.

Sorry -60-70 is obvius, I can't help that.

-70dB is total level (sum) of relvant and irrelevant differences. All distortion
differences must lie inside that -70dB. Treshold of hearing any nonliner
distortions which could possibly be created here is higher (by 10dB) that -70dB.

Correlation is simple -- if sum of varius distortions -- those completely
irrelevant, like miniscule frequency responce differences, (below ±0.1dB -- but
of course way's more -70dB -- +-0.1dB is above 1% level difference which means
above -40dB) as well as relevant (for example harmonic & IM distortion
differences) is at -70dB then all those components, both relevant and irrelevant
are below that.

So this is what ya got? No actual citations of actual scientific
literature that established the actual thresholds of human hearing.

List of literature was presented here, I'm not going to repeat it once more.

No
actual measurements for any real world commercial amplifiers.

There were actual measurements in posts by me, Arny and otheres, and you were
replying to those.

Nothing
that is varifiable.

LOL!

I'd like to see this pass peer review. LOL

And back to the beginning... It's repetitive discussion on your part.

Ya still haven't brought the science.

Execpt I did.

and clearly it isn't coming.

It already came.

If
you had it you would have brought it by now.

So I did.

So I am done responding
to your posts until you actually bring something more to the table in
the way of actual scientific support for the belief in amplifier
transparency.

All what was needed was presented. I'm not going to repeat it more times. Use
your newsgroup client or just use google groups.

And, as I wrote before, discussion for a sake of discussion simply does not
interest me.

rgds
\SK
--
"Never underestimate the power of human stupidity" -- L. Lang
--
http://www.tajga.org -- (some photos from my travels)

[Scott[_6_]]

On May 11, 6:41=A0am, Sebastian Kaliszewski
wrote:
I said in my last post that I would only respond to any actual science
you brough to the conversation. I have decided to let the rest of your
new comments stand on their own to demonstrate the utter lack of audio
content not to mention the utter failure to bring any actual science
to the conversation about the scientific support for the belief in
amplifier transparency. All the obfusecation, ad hominem and other
logical fallacies speak for themselves so I will let them stand on
their own sans comment. I leave them in so as to not be accused of
snipping or ignoring them. I am only responding because you did
actually cite one reference so I will respond to that.

Short memory, Scott?

Who wrote: "Nice stack of irrelevant material only some of which would be

considered peer reviewed science."?

Except it did

This one seemed to be obvious to anyone...

Fletcher, H. and Munson, W.A. Loudness, its definition, measurement and
calculation. Journal of the Acoustic Society of America 5, 82-108 (1933)

Alas an actual citation of some real science. OK now lets take this
papaer and it's contents and now show a corolation to amplifier
transparency. Please show how the contents of this research supports
the belief in amplifier transparency. Clearly it does not in any way
directly say anything on the subject of amplifier transparency. So you
still need to show how this supports that belief.


So what? That's exactly your misrepresentation (now clearly stated) of mi=
ne or
others position... could there be amplifiers which are not transparent? O=
f
course, there certainly could be, gosh, some SET stuff has distortions in=
the
range of 10%.


Nope, I do not. But here we're discussion Carver challenge *you* introduc=
ed into
this thread. Carver challenge demonstrates just that it's possible to mak=
e two
completely different amps sound the same.

Otherwise, Carver challenge has little to do to amp transparency, but it =
was
neither me, nor anyone else but *you* who introduced that. You introduced=
that
to create a strawman, to sidetrack the discussion or what?


Really? Carver challenge was test of two amplifiers:
Conrad-Johnson Premier Four & Carvers prototype one.
Now you introduce also Carver's final product. But OK.


Err, what? At least two of them were similar down to -70dB.


Err, where was the claim of transparency?

Strawman, again.

See above few paragprahs of yours.

There you have it. Nothing new. No citations of actual scientific
literature that establishes the thresholds of human hearing. No new
peer reviewed published papers with listening tests of amplifiers. No
varifiable measurments of actual amplifiers driving a real world
speaker load. Nothing. Oh we do have an acknowledgment that SETs may
actually not be transparent.

I will, in conclusion, once again concede a point. All amplifiers are
transparent except the ones that are not.
If any of you want to wave the science flag again be prepared with
some actual science to back it up or be prepared to be called on it.
Unless you can bring something new in the way of actual scientific
support for the belief in amplifier transparency or unless you can
show the corolation between that belief and the one paper you have
cited I think we are actually done here.

[Arny Krueger]

"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.

-80 dB, or "The -80 dB rule" which says that even when "only" -80 dB,
virtually all real-world aritfacts are inaudible.

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.

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 .

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.

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.

[Sebastian Kaliszewski]

Scott wrote:
On May 11, 6:41 am, Sebastian Kaliszewski
wrote:
I said in my last post that I would only respond to any actual science
you brough to the conversation. I have decided to let the rest of your
new comments stand on their own to demonstrate the utter lack of audio
content not to mention the utter failure to bring any actual science
to the conversation about the scientific support for the belief in
amplifier transparency. All the obfusecation, ad hominem and other
logical fallacies speak for themselves so I will let them stand on
their own sans comment. I leave them in so as to not be accused of
snipping or ignoring them.

You arbitrarily cut large part of my comments while leaving other part but
stripped context. I wonder how moderators let it by, since this is obvious and
explicit attempt at misrepresentation. Hence I hope this (last addressed to you)
reply, mainly pointing that misrepresentation, will be accepted as well.


I am only responding because you did
actually cite one reference so I will respond to that.

Short memory, Scott?

Who wrote: "Nice stack of irrelevant material only some of which would be

considered peer reviewed science."?

Except it did

This one seemed to be obvious to anyone...

Fletcher, H. and Munson, W.A. Loudness, its definition, measurement and
calculation. Journal of the Acoustic Society of America 5, 82-108 (1933)

Alas an actual citation of some real science. OK now lets take this
papaer and it's contents and now show a corolation to amplifier
transparency. Please show how the contents of this research supports
the belief in amplifier transparency. Clearly it does not in any way
directly say anything on the subject of amplifier transparency. So you
still need to show how this supports that belief.


Mechanism of that connection has been discussed and explained ad nauseum in this
very thread. That you don't want to get it is not my problem.


[...arbitrarily selection of comments of mine, deprived of any context --
snipped...]
There you have it. Nothing new. No citations of actual scientific
literature that establishes the thresholds of human hearing. No new
peer reviewed published papers with listening tests of amplifiers. No
varifiable measurments of actual amplifiers driving a real world
speaker load. Nothing. Oh we do have an acknowledgment that SETs may
actually not be transparent.

Those comments of mine you arbitrarily filtered and stripped off context were
exactly about Carver challenge. So this is plain and deliberate attempt at
obfuscation and twisting context to suit you.

[...]

To summarise...

You falsely claimed (in parts which you snipped) there were no verifiable data
posted (while you replied to the very messages containing those data with proper
attribution of sources). You also misrepresented others opionions, blatantly
snipped context attemtpint to change their meaning. Thus further discussion with
you makes no sense.

\SK
--
"Never underestimate the power of human stupidity" -- L. Lang
--
http://www.tajga.org -- (some photos from my travels)




[ This part of the thread is closed. -- dsr ]

[Audio Empire]

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.