"Thomas A" wrote in message
om
"Arny Krueger" wrote in message
...
"Thomas A" wrote in message
om
"Arny Krueger" wrote in message
...
"Thomas A" wrote in message
om
"Arny Krueger" wrote in message
...
"Robert Morein" wrote in message

"Arny Krueger" wrote in message
...

Audio Engineering Society Convention Paper 5876: Perceptual
Discrimination between Musical Sounds with and without Very
High Frequency Components Given at the 115th AES Convention in
New York about a month ago. This paper can be ordered from the
AES web site: www.aes.org .

Arny, since I don't have the paper, does it state how and to
what extent high frequency sounds were captured in the program
material?

Yes, it makes that very clear. The paper includes 10 spectral
plots that clearly describe the extent of the high frequency
sounds that were captured in the program material.

How about beat waves and Tartini tones? How high up in frequency
are they audible?

I don't know if its an artifact but here are two sinus tones
recorded at 20 kHz and 21 kHz:

http://hem.bredband.net/b113928/Tartini.wav

Why theorize? Why not just fire up your 24/96 sound card and listen
to the files you can easily download from
http://www.pcabx.com/technical/sample_rates/index.htm ?

Well since I don't have access to 24/96 I'm asking if such
experiments have been done.

Based on download statistics, thousands of them over the past 3
years.

I hear a 1 kHz tone in my example with 20 kHz and
21 kHz, so what happens with tones of e.g. 29 khz and 30 kHz or
higher?


Your milage may vary depending on a number of other variables.
However the effect is irrelevant to the experience of listening to
music for pleasure..

Any research that shows that the 1 kHz tone is audible at
high-freqeuncy and what level of the fundamental tones are needed?

The level of high frequency energy required to stimulate
nonlinearities in order to cause spurious responses in the audible
range to be heard with typical music is too high.

With complex music the Tartinis are probably masked by lower
fundamentals, but my question related to audibility of pure tones.

The levels required are irrelevant to the experience of listening to
music.


Yes probably so, but since it is impossible to prove with 100%
certainty that it is inaudible with every possible music signal
available, one can argue that the frequency limits should be set where
they are inaudible from the true source, whatever the signal is.

It can be so argued, and the argument can be reasonably be rejected on the
grounds that it is unreasonable. No bridge is built to handle the largest
conceivable load because were that done as a rule, very few bridges would
ever get built. Bridges are built to handle the loads that are reasonable to
carry across bridges based on practical experience.

The good news for you is that just about every computer audio interface
except for bottom-feeder specials now has some kind of support for 24/96.
This is like have a bridge can carry the largest conceivable load on tap,
for instant use when you have the disk space and processing time to spare or
waste.

Arny,

No bridge is built to handle the largest conceivable load because were
that done as a rule, very few bridges would ever get built.

Perfect analogy. Further, I don't WANT to hear beat tones, and I don't see
how that makes a musical experience any more enjoyable.

And so many other things are far more important. I chuckle when people argue
about stuff like jitter that's 120 dB below the music, when the low
frequency response in their listening room varies by 20 dB or more all over
the place.

--Ethan

Arny,

No bridge is built to handle the largest conceivable load because were
that done as a rule, very few bridges would ever get built.

Perfect analogy. Further, I don't WANT to hear beat tones, and I don't see
how that makes a musical experience any more enjoyable.

And so many other things are far more important. I chuckle when people argue
about stuff like jitter that's 120 dB below the music, when the low
frequency response in their listening room varies by 20 dB or more all over
the place.

--Ethan

"Ethan Winer" ethanw at ethanwiner dot com wrote in message

Arny,

No bridge is built to handle the largest conceivable load because were
that done as a rule, very few bridges would ever get built.

Perfect analogy. Further, I don't WANT to hear beat tones, and I
don't see how that makes a musical experience any more enjoyable.

Agreed. The ear nonlinearities that lead to tartini tones audible also cause
choral music to sound like it is "shattering" when its too loud. I have
never heard anybody intentionally turn choral music up to hear how well it
"shatters" in order to improve their musical pleasure.

And so many other things are far more important. I chuckle when
people argue about stuff like jitter that's 120 dB below the music,
when the low frequency response in their listening room varies by 20
dB or more all over the place.

Perhaps the biggest irony is the fact that some of the biggest whiners about
jitter were vinyl bigots, who were day-in and day-out deifying FM distortion
(jitter) at least 20-40 dB higher than has ever been seen in any
halfways-decent digital product, and at jitter frequencies where it is more
audible.

One thing about room-related frequency response variations is that if they
are nominal, don't do a very good job of impairing our ability to hear
frequency response variations elsewhere in the chain. Also, some of the same
kinds of architecturally-related frequency response variations affect live
performances. You get a just as nice of a floor bounce from a cello as a
speaker.

"Ethan Winer" ethanw at ethanwiner dot com wrote in message

Arny,

No bridge is built to handle the largest conceivable load because were
that done as a rule, very few bridges would ever get built.

Perfect analogy. Further, I don't WANT to hear beat tones, and I
don't see how that makes a musical experience any more enjoyable.

Agreed. The ear nonlinearities that lead to tartini tones audible also cause
choral music to sound like it is "shattering" when its too loud. I have
never heard anybody intentionally turn choral music up to hear how well it
"shatters" in order to improve their musical pleasure.

And so many other things are far more important. I chuckle when
people argue about stuff like jitter that's 120 dB below the music,
when the low frequency response in their listening room varies by 20
dB or more all over the place.

Perhaps the biggest irony is the fact that some of the biggest whiners about
jitter were vinyl bigots, who were day-in and day-out deifying FM distortion
(jitter) at least 20-40 dB higher than has ever been seen in any
halfways-decent digital product, and at jitter frequencies where it is more
audible.

One thing about room-related frequency response variations is that if they
are nominal, don't do a very good job of impairing our ability to hear
frequency response variations elsewhere in the chain. Also, some of the same
kinds of architecturally-related frequency response variations affect live
performances. You get a just as nice of a floor bounce from a cello as a
speaker.

(Thomas A) wrote in message . com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the 15
KHz tone in one speaker, and the 21 KHz tone at the other speaker, you
may get different results, because this time speaker intermodulation
won't happen.

(Thomas A) wrote in message . com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the 15
KHz tone in one speaker, and the 21 KHz tone at the other speaker, you
may get different results, because this time speaker intermodulation
won't happen.

(KikeG) wrote in message . com...
(Thomas A) wrote in message . com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the 15
KHz tone in one speaker, and the 21 KHz tone at the other speaker, you
may get different results, because this time speaker intermodulation
won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


T

(KikeG) wrote in message . com...
(Thomas A) wrote in message . com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the 15
KHz tone in one speaker, and the 21 KHz tone at the other speaker, you
may get different results, because this time speaker intermodulation
won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


T

"Thomas A" wrote in message
om
(KikeG) wrote in message
. com...
(Thomas A) wrote in message
. com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The
15 + 21 kHz is percieved "smoother" even if played at much higher
volume than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the
15 KHz tone in one speaker, and the 21 KHz tone at the other
speaker, you may get different results, because this time speaker
intermodulation won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


It can be the listener's ears, it can be the power amp, it can be the
speaker. Three possible diagnoses...

KikeG described the usual means used to isolate problems due to speakers and
amps.

"Thomas A" wrote in message
om
(KikeG) wrote in message
. com...
(Thomas A) wrote in message
. com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The
15 + 21 kHz is percieved "smoother" even if played at much higher
volume than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the
15 KHz tone in one speaker, and the 21 KHz tone at the other
speaker, you may get different results, because this time speaker
intermodulation won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


It can be the listener's ears, it can be the power amp, it can be the
speaker. Three possible diagnoses...

KikeG described the usual means used to isolate problems due to speakers and
amps.

Arny,

Perhaps the biggest irony is the fact that some of the biggest whiners
about jitter were vinyl bigots

Bigots, yeah, that's the right word.

A few years ago Eddie Cilleti, who writes tech articles for Mix magazine,
explained what he believes is the reason some people prefer vinyl over CDs.
According to Eddie, vinyl recordings almost always employ high-frequency
compression. Recording lathe cutter heads are very expensive and can be
destroyed if too much high frequency content is applied, so a limiter that
controls only the high frequencies is generally put in the chain. And the
high frequency limiting is what makes music recorded on vinyl sound
"smoother." This theory sounds reasonable to me, and surely makes more sense
than the notion that vinyl is somehow "better" than digital recording or
analog tape, or the surprisingly common belief that science hasn't yet
learned what needs to be measured.

room-related frequency response variations ... don't do a very good job of
impairing our ability to hear frequency response variations elsewhere in the
chain.

Yes, I agree. My point was that people are swatting at flies while ignoring
a herd of elephants stampeding toward them!

--Ethan

Arny,

Perhaps the biggest irony is the fact that some of the biggest whiners
about jitter were vinyl bigots

Bigots, yeah, that's the right word.

A few years ago Eddie Cilleti, who writes tech articles for Mix magazine,
explained what he believes is the reason some people prefer vinyl over CDs.
According to Eddie, vinyl recordings almost always employ high-frequency
compression. Recording lathe cutter heads are very expensive and can be
destroyed if too much high frequency content is applied, so a limiter that
controls only the high frequencies is generally put in the chain. And the
high frequency limiting is what makes music recorded on vinyl sound
"smoother." This theory sounds reasonable to me, and surely makes more sense
than the notion that vinyl is somehow "better" than digital recording or
analog tape, or the surprisingly common belief that science hasn't yet
learned what needs to be measured.

room-related frequency response variations ... don't do a very good job of
impairing our ability to hear frequency response variations elsewhere in the
chain.

Yes, I agree. My point was that people are swatting at flies while ignoring
a herd of elephants stampeding toward them!

--Ethan

"Arny Krueger" wrote in message ...
"Thomas A" wrote in message
om
(KikeG) wrote in message
. com...
(Thomas A) wrote in message
. com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The
15 + 21 kHz is percieved "smoother" even if played at much higher
volume than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the
15 KHz tone in one speaker, and the 21 KHz tone at the other
speaker, you may get different results, because this time speaker
intermodulation won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


It can be the listener's ears, it can be the power amp, it can be the
speaker. Three possible diagnoses...

KikeG described the usual means used to isolate problems due to speakers and
amps.

Arny,

I do agree that the 16/44.1 is enough for practical use, don't
misunderstand me in this. In the end, the quality is set at
recording/mastering and how our speakers and rooms are working
together.

The question I ask is whether something is audible with respect to
beat waves or not and at what levels (e.g. research that have
established any dB levels for any combination of two freqeuncies where
something is audible or not). The technique for using 24/96 is not by
any means as expensive as always build bridges that than hold any
load, so I don't agree that those examples are comparable. Sure I can
try to do the experiments myself as soon as I get something in 24/96
and headphones that go high in frequency. But again, if it has been
done peer reviewed, I would really want to read the results. It's just
a special interest from my side, nothing else.

T

"Arny Krueger" wrote in message ...
"Thomas A" wrote in message
om
(KikeG) wrote in message
. com...
(Thomas A) wrote in message
. com...

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The
15 + 21 kHz is percieved "smoother" even if played at much higher
volume than the single 15 kHz tone in my ears.


You must make sure that the intermodulation is happening just at your
ears, not at your speakers. If you repeat the test, but playing the
15 KHz tone in one speaker, and the 21 KHz tone at the other
speaker, you may get different results, because this time speaker
intermodulation won't happen.

It is well known that the beatwaves gives audible difference tone
distorsion and its part of what we hear in music (eg two flutes
playing quite loud).

http://www.societymusictheory.org/ww....smt/0102.html


It can be the listener's ears, it can be the power amp, it can be the
speaker. Three possible diagnoses...

KikeG described the usual means used to isolate problems due to speakers and
amps.

Arny,

I do agree that the 16/44.1 is enough for practical use, don't
misunderstand me in this. In the end, the quality is set at
recording/mastering and how our speakers and rooms are working
together.

The question I ask is whether something is audible with respect to
beat waves or not and at what levels (e.g. research that have
established any dB levels for any combination of two freqeuncies where
something is audible or not). The technique for using 24/96 is not by
any means as expensive as always build bridges that than hold any
load, so I don't agree that those examples are comparable. Sure I can
try to do the experiments myself as soon as I get something in 24/96
and headphones that go high in frequency. But again, if it has been
done peer reviewed, I would really want to read the results. It's just
a special interest from my side, nothing else.

T

(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

(KikeG) wrote in message om...
(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I am aware of the difference in level. Therefore I have tested them
with different volume settings, and it is still the same impression in
my ears.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

Yes I will do this, no problem.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

How high? How high levels are needed for the so-called HSS?

http://www.acoustics.org/press/133rd/2pea.html


About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

Thanks. I will read it.

(KikeG) wrote in message om...
(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I am aware of the difference in level. Therefore I have tested them
with different volume settings, and it is still the same impression in
my ears.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

Yes I will do this, no problem.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

How high? How high levels are needed for the so-called HSS?

http://www.acoustics.org/press/133rd/2pea.html


About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

Thanks. I will read it.

(KikeG) wrote in message om...
(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

I've read it now. I also made the experiment but very simply played
the signals with the headphones attached near one ear. The effect is
gone, so the intermodulation is caused in the computer sound card!

T
T

(KikeG) wrote in message om...
(Thomas A) wrote in message . com...

Just want to give two more examples:

A pure 15 kHz tone:

http://hem.bredband.net/b113928/15khz.wav

A 15 kHz + 21 kHz tone:

http://hem.bredband.net/b113928/15_21kHz.wav

At high level the beat and Tartini tone can be heard (6 kHz). The 15
+ 21 kHz is percieved "smoother" even if played at much higher volume
than the single 15 kHz tone in my ears.

I downloaded the files, and realized that you results are due to a
flaw in their generation: both files have nearly same peak amplitude,
which necessarily leads to the 15 KHz tone not having same amplitude
in the two files. The 15 KHz has an amplitude of approx. -0.9 dB at
the first file, and an amplitude of approx. -7 dB at the second file.
Thats the source of the 2nd. file sounding smoother, nothing to do
with intermodulation.

I suggest you repeat the test, this time using a separate speaker for
each tone. This way you will both avoid the different amplitudes
issue, and the nonlinearities at your playback chain being the cause
of the intermodulation. You can't use headphones for such a test. But,
when using speakers, don't play your tones very loud, and play them in
short intervals of time, since it's pretty easy to fry speaker
tweeters playing loud steady high frequency tones. I suggest using
short, fade-in, fade-out, test tones.

About Tartini (intermodulation) tones, it's true that they happen
inside the ear. But it's not less true that when one of the tones is
at a ultrasonic frequency, it takes quite a high level of this tone
for the intermodulation effect to appear.

About this, I suggest you read the several times linked
http://world.std.com/~griesngr/intermod.ppt

I've read it now. I also made the experiment but very simply played
the signals with the headphones attached near one ear. The effect is
gone, so the intermodulation is caused in the computer sound card!

T
T