[javawizard]

You hear audiophiles talk about loss of quality with mp3 files. What
is it they're hearing - or not hearing? The biggest issue seems to be
that the maximum frequency that an mp3 can record is 16khz (kilohertz
- cycles per second), but the human ear can hear up to 20khz. Thought
you'd enjoy knowing! - Jeff
www.odd-info.com

[Richard Crowley]

"javawizard" wrote ...
You hear audiophiles talk about loss of quality with mp3 files.
What is it they're hearing - or not hearing?

I suspect that there is no difference between a 24-bit 192K
sample-rate recording of grunge or rap or head-banging metal
"music" and the MP3 compressed version. :-)

OTOH, with more "transparent" music like classical, etc.
it is pretty easy to hear the artifacts of compression.

The biggest issue seems to be that the maximum frequency
that an mp3 can record is 16khz (kilohertz - cycles per second),
but the human ear can hear up to 20khz.

I don't think so. There are several much more obvious deficiencies.
Note further that many people can't hear all the way out to 20KHz
anymore, OTOH, some people here can hear beyond 20KHz.

Just thought you'd like to know.

[Arny Krueger]

"javawizard" wrote in message
oups.com

You hear audiophiles talk about loss of quality with mp3
files. What is it they're hearing - or not hearing? The
biggest issue seems to be that the maximum frequency that
an mp3 can record is 16khz (kilohertz - cycles per
second),

False - many MP3 encoders can set the high frequency limit to be whatever
you want it to be, up to at least 20 KHz.

but the human ear can hear up to 20khz.

Only if you ignore masking, which is what MP3s are all about anyway. If
there is a strong sound anywhere in the approximate 12-20 KHz range, it will
likely mask sounds at higher frequencies.

[[email protected]]

Arny Krueger wrote:
"javawizard" wrote in message
oups.com

You hear audiophiles talk about loss of quality with mp3
files. What is it they're hearing - or not hearing? The
biggest issue seems to be that the maximum frequency that
an mp3 can record is 16khz (kilohertz - cycles per
second),

False - many MP3 encoders can set the high frequency limit to be whatever
you want it to be, up to at least 20 KHz.

but the human ear can hear up to 20khz.

Only if you ignore masking, which is what MP3s are all about anyway. If
there is a strong sound anywhere in the approximate 12-20 KHz range, it will
likely mask sounds at higher frequencies.




If you need lossy compressed, I suggest OGG VORBIS. If you want CD
quality sound, then use lossless compressed FLAC.

[Michael Black]

javawizard ) writes:

This guy seems to be spamming. For the fourth time this year, he hit
another newsgroup with a nominally on topic post, ending in that
pointer to that website. He's either posted the same message to
that newsgroup all four times, or at least the same basic post.
He's not participating in the discussion, he just posts and isn't
heard from until he posts the same thing later.

Michael

[MRC01]

On Sep 20, 9:50 am, "Richard Crowley" wrote:
I don't think so. There are several much more obvious deficiencies.
Note further that many people can't hear all the way out to 20KHz
anymore, OTOH, some people here can hear beyond 20KHz.

First, javawizard's assertion that MP3 cuts off at 16 kHz is wrong.
You can configure a good encoder to go as high as you like. Many MP3
encoders do cut off at 16 kHz but that is not a flaw in MP3, it is a
flawed implementation of it.

As for the more obvious deficiencies, one that I've encountered more
than once is a phasey kind of distorted sound, especially on VBR but
often on CBR too.

But that said, the encoder is just as important as the bit rate. A
poor encoder will sound perhaps less poor at 320k than at 128k but
still poor. A good encoder can be damn close to imperceptible. Try
using LAME 3.97 at its highest quality settings for 320k and VBR.
Very, very few people could hear a difference, and that includes
trained listeners. MP3 as we typically hear it has poor sound quality,
but at its best it can be excellent. I think MP3 gets a bad reputation
because there are so many poor encoders and low bit rates are commonly
used.

The test I have done is to MP3 encode an original WAV file, then
transform the MP3 into another WAV file, so you have two WAV files of
the same material that are perfectly level matched. Then perform a
double blind test. At best, use ABX software on a PC. Alternately,
burn the tracks to a CD and come up with a way to double blind test it
using your best stereo system. Get a helper to swap tracks for you,
put your CD player on random repeat, or whatever.

With a very good MP3 encoder this can be a humbling experience. I can
consistently tell the difference at or near the highest quality
settings, but only on a select number of my absolute best recordings
with extremely careful listening and even then it is very very
difficult. For most material - the vast majority of recordings that
are out there - the highest quality MP3 is subjectively transparent.
To put it into perspective, a change to room setup or speaker
positioning, has a FAR FAR bigger impact on sound than a good MP3
encoding. Even a change in a component such as a preamp, amp or CD
player, while more subtle than room or speaker changes, can often be
more easily heard than a high quality MP3 encoding.

If anyone wants, I can post links to two WAV files, one MP3 encoded,
and see if anyone can tell which is the original. Might be a fun
experiment.

[daz.diamond]

If anyone wants, I can post links to two WAV files, one MP3 encoded,
and see if anyone can tell which is the original. Might be a fun
experiment.

how about posting the difference in a third file ?

daz
xxx

[Geoff]

javawizard wrote:
You hear audiophiles talk about loss of quality with mp3 files. What
is it they're hearing - or not hearing? The biggest issue seems to be
that the maximum frequency that an mp3 can record is 16khz (kilohertz
- cycles per second), but the human ear can hear up to 20khz. Thought
you'd enjoy knowing! - Jeff

No, not just audiophiles, and a hi frequency limit is not the problem.

The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.

geoff

[MRC01]

On Sep 20, 3:07 pm, "Geoff" wrote:
The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.

Yes, let's do that. They're about 3 MB each. Please don't slam my site
or I'll have to take them down - thanks. Feel free to upload them to a
higher bandwidth more reliable host.
http://mclements.net/Mike/MP3/TestA.wav
http://mclements.net/Mike/MP3/TestB.wav
Which is the original? Anyone can tell if you analyze the waveform.
The question is, can you tell by listening in a double blind test? For
me, it is possible... but not easy. MP3 isn't transparent, but at high
bit rates with a good encoder its artifacts are far from obvious. To
me it seems much better than people seem to think.

[Randy Yates]

MRC01 writes:

On Sep 20, 3:07 pm, "Geoff" wrote:
The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.

Yes, let's do that. They're about 3 MB each. Please don't slam my site
or I'll have to take them down - thanks. Feel free to upload them to a
higher bandwidth more reliable host.
http://mclements.net/Mike/MP3/TestA.wav
http://mclements.net/Mike/MP3/TestB.wav
Which is the original? Anyone can tell if you analyze the waveform.
The question is, can you tell by listening in a double blind test? For
me, it is possible... but not easy. MP3 isn't transparent, but at high
bit rates with a good encoder its artifacts are far from obvious. To
me it seems much better than people seem to think.

TestA.wav sounds very slightly better to me than TestB.wav. Seems like
B is muffled just a bit.

I agree, however, that MP3's - at least the ones I've tried to compare - are
very very close to CD - most of the time indistinguishable for me.
--
% Randy Yates % "So now it's getting late,
%% Fuquay-Varina, NC % and those who hesitate
%%% 919-577-9882 % got no one..."
%%%% % 'Waterfall', *Face The Music*, ELO
http://home.earthlink.net/~yatescr

[Steven Sullivan]

Richard Crowley wrote:
"javawizard" wrote ...
You hear audiophiles talk about loss of quality with mp3 files.
What is it they're hearing - or not hearing?

I suspect that there is no difference between a 24-bit 192K
sample-rate recording of grunge or rap or head-banging metal
"music" and the MP3 compressed version. :-)

OTOH, with more "transparent" music like classical, etc.
it is pretty easy to hear the artifacts of compression.

The biggest issue seems to be that the maximum frequency
that an mp3 can record is 16khz (kilohertz - cycles per second),
but the human ear can hear up to 20khz.

I don't think so. There are several much more obvious deficiencies.

Actually, it's anything but obvous. Most people are extremely hard
put to pass a blind test of a good mp3 vs source. And mp3s certainly
do not have a 16kHz limit



___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

daz.diamond wrote:
If anyone wants, I can post links to two WAV files, one MP3 encoded,
and see if anyone can tell which is the original. Might be a fun
experiment.

how about posting the difference in a third file ?

Listening to the 'difference' by
itself is not the same as listening to the whole files.


___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

Geoff wrote:
javawizard wrote:
You hear audiophiles talk about loss of quality with mp3 files. What
is it they're hearing - or not hearing? The biggest issue seems to be
that the maximum frequency that an mp3 can record is 16khz (kilohertz
- cycles per second), but the human ear can hear up to 20khz. Thought
you'd enjoy knowing! - Jeff

No, not just audiophiles, and a hi frequency limit is not the problem.

The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.

yes, do. when that's been done in *blind* tests, most people can't
tell them apart, when the mp3 is well-made.




___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Steven Sullivan]

MRC01 wrote:
On Sep 20, 3:07 pm, "Geoff" wrote:
The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.

Yes, let's do that. They're about 3 MB each.

what encoder, and waht settings?


___
-S
"As human beings, we understand the world through simile, analogy,
metaphor, narrative and, sometimes, claymation." - B. Mason

[Mr.T]

"Richard Crowley" wrote in message
...
Note further that many people can't hear all the way out to 20KHz
anymore, OTOH, some people here can hear beyond 20KHz.

Well some people can definitely hear above 20 kHz, mainly young girls.
OTOH I can't say I've seen many posts from young girls here on r.a.t, have
you?

MrT.

[Mr.T]

"MRC01" wrote in message
ups.com...
First, javawizard's assertion that MP3 cuts off at 16 kHz is wrong.
You can configure a good encoder to go as high as you like. Many MP3
encoders do cut off at 16 kHz but that is not a flaw in MP3, it is a
flawed implementation of it.

No it's a valid decision on what to leave out in order to reduce the bit
rate to that required.
SOMETHING must be left out in any lossy compression scheme, and it's not
just the high frequencies. However those are usually the least objectionable
as first choice.

MrT.

[MRC01]

On Sep 20, 7:33 pm, Steven Sullivan wrote:
MRC01 wrote:
On Sep 20, 3:07 pm, "Geoff" wrote:
The prolem is that much detail is lost, and sometimes even insignificant
parts are boosted. Subtle detail to sounds such as harmonics and
reverberation can either get totally lost, or turned into a phasey swoosh.

Try it yourself. Rip a CD track and encode it at different rates to MP3.
Then burn the MP3s back to an audio CD along with the original track.
Compare away.
Yes, let's do that. They're about 3 MB each.

what encoder, and waht settings?

LAME 3.97, --cbr -b 320 -q0 -m s -k

[MRC01]

On Sep 20, 7:33 pm, Steven Sullivan wrote:
what encoder, and waht settings?

One of the above is encoded using LAME 3.97: --cbr -b 320 -q0 -m s -k

I also tried: --vbr-old -V0 -q0 -m s -k. This makes a smaller file but
the differences are ever so slightly more perceptible. Maybe.
Theoretically, VBR gives each frame whatever bit rate is necessary to
encode the signal at a consistent quality. This gives significantly
smaller file sizes with virtually no difference in quality compared to
CBR 320. LAME is one of the few encoders that actually achieves this.
In practice, many MP3 encoders absolutely suck at VBR, producing
audible artifacts at all bit rates. Also, some players don't handle
VBR well. To keep things simple I stuck with CBR for this test.

[MRC01]

On Sep 20, 8:29 pm, "Mr.T" MrT@home wrote:
"MRC01" wrote in message

ups.com...

First, javawizard's assertion that MP3 cuts off at 16 kHz is wrong.
You can configure a good encoder to go as high as you like. Many MP3
encoders do cut off at 16 kHz but that is not a flaw in MP3, it is a
flawed implementation of it.

No it's a valid decision on what to leave out in order to reduce the bit
rate to that required.
SOMETHING must be left out in any lossy compression scheme, and it's not
just the high frequencies. However those are usually the least objectionable
as first choice.

Sure, at lower bit rates, especially since lower frequencies are
likely to mask the higher ones. But I've seen encoders that cut off at
16k even at the highest bit rates. This IMO is a flaw. Ideally, the
cutoff should either be user configurable, or it should depend on the
bit rate and the material being encoded.

[daz.diamond]

Steven Sullivan wrote:
daz.diamond wrote:
If anyone wants, I can post links to two WAV files, one MP3 encoded,
and see if anyone can tell which is the original. Might be a fun
experiment.

how about posting the difference in a third file ?

Listening to the 'difference' by
itself is not the same as listening to the whole files.

that's not the point - hence _third_ file

1 : hear wav
2 : hear mp3
3 : evaluate _actual_ difference compared to what you perceived

daz

[Mr.T]

"MRC01" wrote in message
ups.com...
First, javawizard's assertion that MP3 cuts off at 16 kHz is wrong.
You can configure a good encoder to go as high as you like. Many MP3
encoders do cut off at 16 kHz but that is not a flaw in MP3, it is a
flawed implementation of it.

No it's a valid decision on what to leave out in order to reduce the bit
rate to that required.
SOMETHING must be left out in any lossy compression scheme, and it's not
just the high frequencies. However those are usually the least
objectionable
as first choice.

Sure, at lower bit rates, especially since lower frequencies are
likely to mask the higher ones. But I've seen encoders that cut off at
16k even at the highest bit rates. This IMO is a flaw.

Absolutely, if by higher bit rates you mean ~1.4Mb/s :-)


Ideally, the cutoff should either be user configurable,

No argument from me.

or it should depend on the bit rate and the material being encoded.

No, I'd stick with user configureable.

MrT.

[MRC01]

On Sep 20, 3:38 pm, Randy Yates wrote:
MRC01 writes:

Yes, let's do that. They're about 3 MB each. Please don't slam my site
or I'll have to take them down - thanks. Feel free to upload them to a
higher bandwidth more reliable host.
http://mclements.net/Mike/MP3/TestA.wav
http://mclements.net/Mike/MP3/TestB.wav
Which is the original? Anyone can tell if you analyze the waveform.
The question is, can you tell by listening in a double blind test? For
me, it is possible... but not easy. MP3 isn't transparent, but at high
bit rates with a good encoder its artifacts are far from obvious. To
me it seems much better than people seem to think.

TestA.wav sounds very slightly better to me than TestB.wav. Seems like
B is muffled just a bit.

I agree, however, that MP3's - at least the ones I've tried to compare - are
very very close to CD - most of the time indistinguishable for me.

Just FYI, now that people have had a chance to listen, TestA is the
MP3 version, TestB is the original. This is easy to see from a
spectrum analysis. The original has energy all the way to and beyond
20 kHz while TestA has a LP filter between 19 and 20 kHz. But it's
much more difficult to hear this in a DBT.

I can see from my site logs that many people downloaded the clips,
though Randy was the only one brave enough to venture an opinion.