On Tue, 17 Feb 2004 06:11:03 -0000, Rich Andrews
wrote:

Randy Yates wrote in :

(Svante) writes:

Randy Yates wrote in message news:
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Svante, by all means, I appreciate your response. This is good
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

That it is lossy compression, and therefore *measurably* diffferent,
does not mean that it's *audibly* different. Only *listening* tests,
under controlled conditions, can determine audibility. General
repoirts seem to indiucate that 128kbs is audible, but 192 may be
adequate for all but the most critical applications.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

"citronzx" wrote in message link.net...
Svante, what do you teach?

Mainly electroacoustics at the Royal Institute of Technology, Sweden.
But this is another course that is actually about project control, and
I supply them with the task of determining the audibility of mp3s.


"Svante" wrote in message
om...
Randy Yates wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

"citronzx" wrote in message link.net...
Svante, what do you teach?

Mainly electroacoustics at the Royal Institute of Technology, Sweden.
But this is another course that is actually about project control, and
I supply them with the task of determining the audibility of mp3s.


"Svante" wrote in message
om...
Randy Yates wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

"citronzx" wrote in message link.net...
Svante, what do you teach?

Mainly electroacoustics at the Royal Institute of Technology, Sweden.
But this is another course that is actually about project control, and
I supply them with the task of determining the audibility of mp3s.


"Svante" wrote in message
om...
Randy Yates wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

"citronzx" wrote in message link.net...
Svante, what do you teach?

Mainly electroacoustics at the Royal Institute of Technology, Sweden.
But this is another course that is actually about project control, and
I supply them with the task of determining the audibility of mp3s.


"Svante" wrote in message
om...
Randy Yates wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Rich Andrews writes:
Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

Thanks for your input, Rich.
--
% Randy Yates % "Watching all the days go by...
%% Fuquay-Varina, NC % Who are you and who am I?"
%%% 919-577-9882 % 'Mission (A World Record)',
%%%% % *A New World Record*, ELO
http://home.earthlink.net/~yatescr

Rich Andrews writes:
Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

Thanks for your input, Rich.
--
% Randy Yates % "Watching all the days go by...
%% Fuquay-Varina, NC % Who are you and who am I?"
%%% 919-577-9882 % 'Mission (A World Record)',
%%%% % *A New World Record*, ELO
http://home.earthlink.net/~yatescr

Rich Andrews writes:
Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

Thanks for your input, Rich.
--
% Randy Yates % "Watching all the days go by...
%% Fuquay-Varina, NC % Who are you and who am I?"
%%% 919-577-9882 % 'Mission (A World Record)',
%%%% % *A New World Record*, ELO
http://home.earthlink.net/~yatescr

Rich Andrews writes:
Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

Thanks for your input, Rich.
--
% Randy Yates % "Watching all the days go by...
%% Fuquay-Varina, NC % Who are you and who am I?"
%%% 919-577-9882 % 'Mission (A World Record)',
%%%% % *A New World Record*, ELO
http://home.earthlink.net/~yatescr

"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.

I think it's more like the difference between Smirnoff versus Absolut.

When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Are you pompous enough to maintain that your subjective impressions are
the reference by which we should all compare? What incredible stupidity.
--
% Randy Yates % "And all that I can do
%% Fuquay-Varina, NC % is say I'm sorry,
%%% 919-577-9882 % that's the way it goes..."
%%%% % Getting To The Point', *Balance of Power*, ELO
http://home.earthlink.net/~yatescr

"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.

I think it's more like the difference between Smirnoff versus Absolut.

When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Are you pompous enough to maintain that your subjective impressions are
the reference by which we should all compare? What incredible stupidity.
--
% Randy Yates % "And all that I can do
%% Fuquay-Varina, NC % is say I'm sorry,
%%% 919-577-9882 % that's the way it goes..."
%%%% % Getting To The Point', *Balance of Power*, ELO
http://home.earthlink.net/~yatescr

"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.

I think it's more like the difference between Smirnoff versus Absolut.

When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Are you pompous enough to maintain that your subjective impressions are
the reference by which we should all compare? What incredible stupidity.
--
% Randy Yates % "And all that I can do
%% Fuquay-Varina, NC % is say I'm sorry,
%%% 919-577-9882 % that's the way it goes..."
%%%% % Getting To The Point', *Balance of Power*, ELO
http://home.earthlink.net/~yatescr

"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.

I think it's more like the difference between Smirnoff versus Absolut.

When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Are you pompous enough to maintain that your subjective impressions are
the reference by which we should all compare? What incredible stupidity.
--
% Randy Yates % "And all that I can do
%% Fuquay-Varina, NC % is say I'm sorry,
%%% 919-577-9882 % that's the way it goes..."
%%%% % Getting To The Point', *Balance of Power*, ELO
http://home.earthlink.net/~yatescr

"Randy Yates" wrote in message

"Arny Krueger" writes:

"Randy Yates" wrote in message


Given a relatively good MP3 encoder, has any of your ABX testing
been able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

As a rule, its possible to hear at last slight differences,
especially if you pick the music to exploit the failings of the
coder.

Thanks Arny. I cannot, even with the proported "difficult" jangling
keys passage - I suppose I'm just getting old.

Or you've got a coder that does well with keys jangling. I know for sure
that people have diddled coders that failed miserably with my test files,
until they ran clean. Say, Microsoft...

The people on the Hydrogen Audio forum like to sift CDs looking for stuff
that breaks coders. They're good people to get the latest-greatest tips
from.

http://www.hydrogenaudio.org/

"Randy Yates" wrote in message

"Arny Krueger" writes:

"Randy Yates" wrote in message


Given a relatively good MP3 encoder, has any of your ABX testing
been able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

As a rule, its possible to hear at last slight differences,
especially if you pick the music to exploit the failings of the
coder.

Thanks Arny. I cannot, even with the proported "difficult" jangling
keys passage - I suppose I'm just getting old.

Or you've got a coder that does well with keys jangling. I know for sure
that people have diddled coders that failed miserably with my test files,
until they ran clean. Say, Microsoft...

The people on the Hydrogen Audio forum like to sift CDs looking for stuff
that breaks coders. They're good people to get the latest-greatest tips
from.

http://www.hydrogenaudio.org/

"Randy Yates" wrote in message

"Arny Krueger" writes:

"Randy Yates" wrote in message


Given a relatively good MP3 encoder, has any of your ABX testing
been able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

As a rule, its possible to hear at last slight differences,
especially if you pick the music to exploit the failings of the
coder.

Thanks Arny. I cannot, even with the proported "difficult" jangling
keys passage - I suppose I'm just getting old.

Or you've got a coder that does well with keys jangling. I know for sure
that people have diddled coders that failed miserably with my test files,
until they ran clean. Say, Microsoft...

The people on the Hydrogen Audio forum like to sift CDs looking for stuff
that breaks coders. They're good people to get the latest-greatest tips
from.

http://www.hydrogenaudio.org/

"Randy Yates" wrote in message

"Arny Krueger" writes:

"Randy Yates" wrote in message


Given a relatively good MP3 encoder, has any of your ABX testing
been able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

As a rule, its possible to hear at last slight differences,
especially if you pick the music to exploit the failings of the
coder.

Thanks Arny. I cannot, even with the proported "difficult" jangling
keys passage - I suppose I'm just getting old.

Or you've got a coder that does well with keys jangling. I know for sure
that people have diddled coders that failed miserably with my test files,
until they ran clean. Say, Microsoft...

The people on the Hydrogen Audio forum like to sift CDs looking for stuff
that breaks coders. They're good people to get the latest-greatest tips
from.

http://www.hydrogenaudio.org/

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that is
inaudible. The only questions remaining are how much and how to.

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that is
inaudible. The only questions remaining are how much and how to.

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that is
inaudible. The only questions remaining are how much and how to.

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that is
inaudible. The only questions remaining are how much and how to.

Stewart Pinkerton wrote in message . ..
On Tue, 17 Feb 2004 06:11:03 -0000, Rich Andrews
wrote:

Randy Yates wrote in :

(Svante) writes:

Randy Yates wrote in message news:
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Svante, by all means, I appreciate your response. This is good
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

That it is lossy compression, and therefore *measurably* diffferent,
does not mean that it's *audibly* different. Only *listening* tests,
under controlled conditions, can determine audibility. General
repoirts seem to indiucate that 128kbs is audible, but 192 may be
adequate for all but the most critical applications.

I agree, with the additional condition that the encoder is good. There
certainly are encoders out there that are audible even at 320 kbit/s
(I just see that I wrote "mbit/s" in my previous post, milli-bit/s,
oops...).

This encoder difference is one of the things that usually comes out of
the students' tests.

Stewart Pinkerton wrote in message . ..
On Tue, 17 Feb 2004 06:11:03 -0000, Rich Andrews
wrote:

Randy Yates wrote in :

(Svante) writes:

Randy Yates wrote in message news:
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Svante, by all means, I appreciate your response. This is good
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

That it is lossy compression, and therefore *measurably* diffferent,
does not mean that it's *audibly* different. Only *listening* tests,
under controlled conditions, can determine audibility. General
repoirts seem to indiucate that 128kbs is audible, but 192 may be
adequate for all but the most critical applications.

I agree, with the additional condition that the encoder is good. There
certainly are encoders out there that are audible even at 320 kbit/s
(I just see that I wrote "mbit/s" in my previous post, milli-bit/s,
oops...).

This encoder difference is one of the things that usually comes out of
the students' tests.

Stewart Pinkerton wrote in message . ..
On Tue, 17 Feb 2004 06:11:03 -0000, Rich Andrews
wrote:

Randy Yates wrote in :

(Svante) writes:

Randy Yates wrote in message news:
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Svante, by all means, I appreciate your response. This is good
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

That it is lossy compression, and therefore *measurably* diffferent,
does not mean that it's *audibly* different. Only *listening* tests,
under controlled conditions, can determine audibility. General
repoirts seem to indiucate that 128kbs is audible, but 192 may be
adequate for all but the most critical applications.

I agree, with the additional condition that the encoder is good. There
certainly are encoders out there that are audible even at 320 kbit/s
(I just see that I wrote "mbit/s" in my previous post, milli-bit/s,
oops...).

This encoder difference is one of the things that usually comes out of
the students' tests.

Stewart Pinkerton wrote in message . ..
On Tue, 17 Feb 2004 06:11:03 -0000, Rich Andrews
wrote:

Randy Yates wrote in :

(Svante) writes:

Randy Yates wrote in message news:
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Pardon me for responding to a question not directed to me, but I think
I have some input to this. I have a group of undergraduate students
every year that perform A/B testing on mp3 compression. Their task is
to find compressors on the web, rip a few CDs of their choice and to
run A/B testing with a software supplied by me. The software
randomises the testing, so they can do it without too big risk of
errors. I am not present during the tests, however. 99% confidence is
required (7 correct responses of 7 tries or equivalent).
Last year the students tested three encoders. According to their
report they tested Lame, bladeenc and mpegenc. They tested 8 CDs. At
128 mbit/s and with the lame encoder they managed to detect 3 of these
8. Same number for the other two encoders was 7 of 8. At 160 mbit/s
they still got 99% confidence for 1 of 8 with lame, 2 of 8 with
bladeenc, and 5 of 8 for mpegenc.
Now, this is a student report, so I would not take it as 100%
certainly true, but the blinded A/B procedure makes the test
reasonably well controlled.

Svante, by all means, I appreciate your response. This is good
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

That it is lossy compression, and therefore *measurably* diffferent,
does not mean that it's *audibly* different. Only *listening* tests,
under controlled conditions, can determine audibility. General
repoirts seem to indiucate that 128kbs is audible, but 192 may be
adequate for all but the most critical applications.

I agree, with the additional condition that the encoder is good. There
certainly are encoders out there that are audible even at 320 kbit/s
(I just see that I wrote "mbit/s" in my previous post, milli-bit/s,
oops...).

This encoder difference is one of the things that usually comes out of
the students' tests.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Well, if someone says that it is impossible to differentiate between
vodka and water, the way to *prove* him/her wrong is to do a blind
test. For sure it will succeed. Without the test the difference might
be obvious, but not proven.
Anyone who states that a difference is obvious should not be afrait to
prove it with an A/B test, IMO.

Also, the choice of encoder is important, and while 128 kbit/s appears
to be audible in A/B tests even for the good encoder, it also appears
as if many can live with it.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Well, if someone says that it is impossible to differentiate between
vodka and water, the way to *prove* him/her wrong is to do a blind
test. For sure it will succeed. Without the test the difference might
be obvious, but not proven.
Anyone who states that a difference is obvious should not be afrait to
prove it with an A/B test, IMO.

Also, the choice of encoder is important, and while 128 kbit/s appears
to be audible in A/B tests even for the good encoder, it also appears
as if many can live with it.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Well, if someone says that it is impossible to differentiate between
vodka and water, the way to *prove* him/her wrong is to do a blind
test. For sure it will succeed. Without the test the difference might
be obvious, but not proven.
Anyone who states that a difference is obvious should not be afrait to
prove it with an A/B test, IMO.

Also, the choice of encoder is important, and while 128 kbit/s appears
to be audible in A/B tests even for the good encoder, it also appears
as if many can live with it.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

Well, if someone says that it is impossible to differentiate between
vodka and water, the way to *prove* him/her wrong is to do a blind
test. For sure it will succeed. Without the test the difference might
be obvious, but not proven.
Anyone who states that a difference is obvious should not be afrait to
prove it with an A/B test, IMO.

Also, the choice of encoder is important, and while 128 kbit/s appears
to be audible in A/B tests even for the good encoder, it also appears
as if many can live with it.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.

"Charles Tomaras" wrote in message ...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for any of
the
codecs that I have heard. The differences are pretty obvious to most
astute
listeners even on moderately priced equipment. As a jazz fan I really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test to
tell the difference between vodka and water. I'm sorry if one cannot hear
the difference between Redbook CD and 128kbps material compressed with the
popular codecs.....I think it's as obvious as the water and vodka analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.

(Svante) wrote in
om:

"Charles Tomaras" wrote in message news:
...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing
been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for
any of
the
codecs that I have heard. The differences are pretty obvious to
most
astute
listeners even on moderately priced equipment. As a jazz fan I
really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to
circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test
to
tell the difference between vodka and water. I'm sorry if one cannot
hear
the difference between Redbook CD and 128kbps material compressed with
the
popular codecs.....I think it's as obvious as the water and vodka
analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.


One need not a/b blind taste test water and vodka. The lit match test
works fine.

I think it can be said that obvious differences don't need an elaborate
DBT or anything similar. A blind man may not know if the moon is full or
not, but an elaborate test isn't necessary to find out either.

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

(Svante) wrote in
om:

"Charles Tomaras" wrote in message news:
...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing
been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for
any of
the
codecs that I have heard. The differences are pretty obvious to
most
astute
listeners even on moderately priced equipment. As a jazz fan I
really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to
circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test
to
tell the difference between vodka and water. I'm sorry if one cannot
hear
the difference between Redbook CD and 128kbps material compressed with
the
popular codecs.....I think it's as obvious as the water and vodka
analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.


One need not a/b blind taste test water and vodka. The lit match test
works fine.

I think it can be said that obvious differences don't need an elaborate
DBT or anything similar. A blind man may not know if the moon is full or
not, but an elaborate test isn't necessary to find out either.

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

(Svante) wrote in
om:

"Charles Tomaras" wrote in message news:
...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing
been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for
any of
the
codecs that I have heard. The differences are pretty obvious to
most
astute
listeners even on moderately priced equipment. As a jazz fan I
really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to
circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test
to
tell the difference between vodka and water. I'm sorry if one cannot
hear
the difference between Redbook CD and 128kbps material compressed with
the
popular codecs.....I think it's as obvious as the water and vodka
analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.


One need not a/b blind taste test water and vodka. The lit match test
works fine.

I think it can be said that obvious differences don't need an elaborate
DBT or anything similar. A blind man may not know if the moon is full or
not, but an elaborate test isn't necessary to find out either.

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

(Svante) wrote in
om:

"Charles Tomaras" wrote in message news:
...
"Randy Yates" wrote in message
...
"Charles Tomaras" writes:

"Randy Yates" wrote in message
...
Arny,

Given a relatively good MP3 encoder, has any of your ABX testing
been
able to quantify whether or not people can reliable detect the
difference between a CD recording and 128 kbps MP3?

Doesn't take special tests to hear the difference at 128kbps for
any of
the
codecs that I have heard. The differences are pretty obvious to
most
astute
listeners even on moderately priced equipment. As a jazz fan I
really
hear
it on the high end of cymbals with swirly sort of phase issues.

This is exactly the sort of unscientific assertion I wanted to
circumvent
by my ABX-qualified question.

While I can appreciate the accuracy of blind testing, I recently heard
someone post on a newsgroup that you don't need a blind comparison test
to
tell the difference between vodka and water. I'm sorry if one cannot
hear
the difference between Redbook CD and 128kbps material compressed with
the
popular codecs.....I think it's as obvious as the water and vodka
analogy.
When you get up to 192kbps and above the differences are much less
pronounced but at 128kpbs it's not even close.

As long as someone states that water and vodka taste the same, the way
to prove he/she is wrong is to do the test. The test will obviously
succed, but until it's done, the difference isn't proven.


One need not a/b blind taste test water and vodka. The lit match test
works fine.

I think it can be said that obvious differences don't need an elaborate
DBT or anything similar. A blind man may not know if the moon is full or
not, but an elaborate test isn't necessary to find out either.

r

--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

"Arny Krueger" wrote in news:s9ydnS-SwoGXsK_dRVn-
:

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known
to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that
is
inaudible. The only questions remaining are how much and how to.





I think that the two biggest variables in determining audibility of a
compression scheme is the equipment used in playback and the listeners
ability to hear differences.

r


--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

"Arny Krueger" wrote in news:s9ydnS-SwoGXsK_dRVn-
:

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known
to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that
is
inaudible. The only questions remaining are how much and how to.





I think that the two biggest variables in determining audibility of a
compression scheme is the equipment used in playback and the listeners
ability to hear differences.

r


--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.

"Arny Krueger" wrote in news:s9ydnS-SwoGXsK_dRVn-
:

"Rich Andrews" wrote in message
. 44
Randy Yates wrote in :
information.

Even at 320kc the MP3 is audibly different. It cannot help but be
different. It is a lossy compression.

A form of lossy compression called "14/32 Linear PCM" is generally known
to
be unconditionally transparent. Failing that, kick the numbers up a bit.

The point is that it is humanly possible to lose data in a fashion that
is
inaudible. The only questions remaining are how much and how to.





I think that the two biggest variables in determining audibility of a
compression scheme is the equipment used in playback and the listeners
ability to hear differences.

r


--
Nothing beats the bandwidth of a station wagon filled with DLT tapes.