Reply
 
Thread Tools Display Modes
  #1   Report Post  
Erwin Timmerman
 
Posts: n/a
Default Louder IS Better (With Lossy)

Lord Hasenpfeffer wrote:

I still have to wonder about the "threshold of human hearing" they speak
about, because with playback systems that have a volume knob, it just doesn't
apply.


Unless I am mistaken, frequencies discarded during encoding remain
absent during playback regardless of volume.


Yep. That's why setting an absolute threshold (relative to FS) is strange to me. It
assumes that everybody is listening at the same level, which, with a volume knob,
isn't necessarily the case.

This would mean that indeed if you raise the level of a song, more stuff gets
encoded.

However, Bob has quite a point where he says that encoding more frequencies leaves
less bandwith for encoding the stronger frequencies correctly. This applies
especially to fixed bit rate encoding. Even more so, raising the absolute level of
the *noise* (what you're doing when raising the level of a whole song) will make
the encoder work harder to encode that noise more precisely, thus taking bandwith
away from encoding the real stuff. The whole idea of the lossy compression is to
leave out what you won't hear. Trying to get those frequencies back in goes against
the whole idea of compression, and will hurt the non-left-out frequencies.
Especially when it would be extra noise you're encoding.

The amount of noise, and the way it influences encoding, differs per song and
therefore results will vary between songs. So alas I think when you want to achieve
optimum results it is still a matter of listening, and not setting standard batch
levels for all songs.

Your own preferences OTOH are of course an entirely different matter. If I
understood it correctly, your wish of "normalizing" songs is also fed by the fact
that you want to standardize your listening experience. Nothing wrong with that,
it's just your own preference. Maybe it would be more wise to find out what crest
factor you really like and stick to that (and hope you won't change your preference
after the stuff is encoded), with the listening experience itself in mind and not
the encoding.

FWIW, I'm currently in the process of transferring my own CD collection to MP3 for
playback on the DVD player. The number of CD's I have just takes too much space in
the living room. So I will put the stuff on MP3 CD's for background music and keep
the CD's in the attic for archiving, and get a CD from there when I actually really
have the time to sit down and listen to the music (which will be only a few times a
year I guess).

I rust rip and encode, for two reasons: 1) it doesn't tamper with the music as-is,
2) it takes less time to do. For the encoding I use the lame windows encoder (pun
intended) with the setting VBR, highest quality, 64-320 kb/s. With the 50 CD's I've
already done so far, I didn't notice any very obvious artefacts yet (and I know
what to listen for.. swishy cymbals, strange bass to name a few). That's why I'm
interested in the tracks Peter mentioned, tracks that give encoders a hard time.
Having these disks myself makes it easy to try the test.

If you want to order a set as well to do the same tests with your encoder, you can
find the ordering info at http://www.recaudiopro.net
Ordering one is worth the enjoyment of the tracks, as well as the learning
experience by reading the liner notes: "how did they do that??". Highly
recommended.

Good luck,

Erwin Timmerman.

  #2   Report Post  
Lord Hasenpfeffer
 
Posts: n/a
Default Louder IS Better (With Lossy)

Erwin Timmerman wrote:

I still have to wonder about the "threshold of human hearing" they speak
about, because with playback systems that have a volume knob, it just doesn't
apply.


Unless I am mistaken, frequencies discarded during encoding remain
absent during playback regardless of volume.


Yep. That's why setting an absolute threshold (relative to FS) is strange to me. It
assumes that everybody is listening at the same level, which, with a volume knob,
isn't necessarily the case.


The Absolute Threshold of Hearing is only in effect during *encoding*,
and *not* during playback. It's just a compression scheme's way of
saying, "Well, according to the chart here, *that* frequency's not gonna
be heard so... out it goes! Whoops, there's another one... take that
one out too!"

Which is why I've been saying that it only stands to reason that a
low-level WAV will produce a "bad sounding" MP3 and, unlike with CD,
simply turning up the volume will *not* solve the problem - because the
frequencies are already not there to be heard at any volume. The never
made it across to become a part of the MP3 to begin with.

It was obviously an apparent practice in the 80s to master CDs at much
lower peak *and* RMS levels than is the case today - so naturally if I
want to make "good sounding" MP3s from any of my older, quieter CDs
which were mastered that far back, I'm going to want to boost the
average RMS levels of the WAVs I rip from them before I encode them.

This would mean that indeed if you raise the level of a song, more stuff gets
encoded.


Precisely - at least I *think* so. Which is why this discussion exists.

Yet this is the point at which every seems to want to ignore the topic
at hand and start chiding me for changing the original dynamic range,
etc. - which is all fine and good. I understand and accept what they're
saying even though some of them *still* don't believe that I do and
always have. When I tell them that that's completely unrelated to my
stated mission, however, I think they think I'm declaring them to be
fools and idiots while I'm declaring myself to be the Lord - which isn't
the case at all.

The MFSL CD edition of Pink Floyd's "Dark Side Of The Moon" CD has an
exceedingly low maximum peak level by today's standards which obviously
suck, depending on whom you ask. Since CD audio is uncompressed,
however, this is not a problem. Simply turn up the volume and ta-da!
On the other hand, I believe that that particular CD would make a
horrible MP3 because so many of its low-amp frequencies would simply be
filtered out and never make it to the destination file. Amplifying the
WAV, however, would enable more frequencies to survive the ATH-based
filtration.

What other factors are at play here? CBR is one for which VBR seems to
be a good solution since it theoretically would enable the MP3 to retain
all the extra frequencies which "made it over the hurdle".

raising the absolute level of the *noise* (what you're doing when
raising the level of a whole song) will make the encoder work harder
to encode that noise more precisely, thus taking bandwith away from
encoding the real stuff.


Correct me if I'm wrong, but except in only the quietest passages would
the noise not be "masked" by the real stuff... and masked sounds are
*also* filtered out by separate, additional process unrelated to the
ATH-based filtration.

So alas I think when you want to achieve optimum results it is still a
a matter of listening, and not setting standard batch levels for all
songs.


But the purpose for setting a standard batch level is for raising the
amplitudes of all WAVs which have been ripped from a single common
source. This preserves the *original relative amplitude* of all the
WAVs in question which is highly desireable. You usually would not want
a quiet song and a loud song from the same album to be equalized in
terms of loudness. That would not be pleasant.

Your own preferences OTOH are of course an entirely different matter.
If I understood it correctly, your wish of "normalizing" songs is also
fed by the fact that you want to standardize your listening experience.


Correct.

Nothing wrong with that, it's just your own preference.


That's right. I am not interested in doing "undue adverse harm" to the
dynamic range of a recording for the sake of sheer amplitude.
Personally, when producing MP3s from CDs, I don't care if a little peak
here and a little peak there aren't perfectly preserved at their
original loudnesses if the frequencies throughout the *entire rest of
the recording* are better able to survive the effects of ATH-based lossy
filtration.

Maybe it would be more wise to find out what crest factor you really
like and stick to that (and hope you won't change your preference
after the stuff is encoded), with the listening experience itself
in mind and not the encoding.


I'm not very familiar with the term "crest factor" but I can *kinda*
discern what it means by the contexts in which I've seen it used so far
elsewhere in this thread. Personally I prefer Colgate.

FWIW, I'm currently in the process of transferring my own CD collection to MP3 for
playback on the DVD player. The number of CD's I have just takes too much space in
the living room. So I will put the stuff on MP3 CD's for background music and keep
the CD's in the attic for archiving, and get a CD from there when I actually really
have the time to sit down and listen to the music (which will be only a few times a
year I guess).


You're aim is similar to mine in many ways. But I actually have about
four different MP3-based projects going on right now - each for a
different personal and/or professional reason.

And, oh... I hope it never gets too hot or cold in your attic!

I rust rip and encode, for two reasons: 1) it doesn't tamper with the music as-is,


Well, yeah it will if your original WAVs' amplitudes are too low.
(At least I *think* so anyway).

If you want to order a set as well to do the same tests with your encoder, you can
find the ordering info at http://www.recaudiopro.net
Ordering one is worth the enjoyment of the tracks, as well as the learning
experience by reading the liner notes: "how did they do that??". Highly
recommended.


I will be looking into it. Thanks!

In the meantime, though, "have a gander" and tell me what you see:

http://www.mykec.com/mykec/images/Su...Sunday_012.gif

My unprofessionally trained eyes see a *few* peaks being slightly
limited (primarily at the 3 1/3 minute mark) but not by any noticeably
harmful degree. Meanwhile, I also see *a lot of potential* for a *lot*
of frequencies being spared from the ATH-block chopping block across the
width of the entire rest of the recording.

Myke

--

-================================-
Windows...It's rebootylicious!!!
-================================-

  #3   Report Post  
John LeBlanc
 
Posts: n/a
Default Louder IS Better (With Lossy)


"David Morgan (MAMS)" wrote in message
...

Stop thinking, for a moment, of amplitude - as in the file's RMS power
of being the determining factor. Move your train of thought from the
vertical scale to one of 'depth'. Stop thinking of frequencies as being
louder or softer, but rather as being in front of or behind others. Some
call that a form of masking.

No matter what the amplitude of the .wav file, the encoding process still
looks for frequencies that are potentially hidden *behind* others.


Exactly. THough I'd throw in another factor: length of time. A unit of
Time-Frequency is broken up into bands and the software deals with that.

If a snare hit at 1046Hz (originally ATRAC band 9 1080-1270Hz) occurs at the
same moment as, say, a mallet striking a xylophone at 2637Hz (originally
ATRAC band 10 1270-1480Hz), and the amplitude of the snare hit exceeds that
of the xylophone to the extent determined by the software as threshold, the
frequency band controlled by the software matching to that of the xylophone
hit is given less attention -- masking -- (introducing quantization noise in
the process) for the length of time the snare hit's amplitude controls the
threshold. It's less data for the encoder to produce since it doesn't have
to give much attention to band 10, which makes for a smaller soundfile on
the other end of the encoding process.

When the amplitude moves below the threshold by moving through the
time-frequency slice, we get the ringing tail of the xylophone. Which, we
are told, is perfectly fine anyway because we couldn't really hear the
xylophone strike behind the snare hit. Psychoacoustic bull****, in my
opinion. But it does help to understand why some audio files get converted
in a more acceptable manner than others. It's all in the timing.

More bands, the center frequency and width of which being adaptive from one
time-frequency slice to another, still removes the subtlety a recording and
mixing engineer could have put in there to begin with. Is it acceptable to
screw up audio in this manner? Evidently the overwhelming response from MP3
fans indicates that the answer is, "Yes." I happen to disagree, but then I
represent just one man's opinion.

Lowering the bar to such a degree (actually, to any degree) for acceptable
audio quality -- 128KHz sample rate MP3 -- is such an odd thing given the
lengths to which some engineers and equipment manufacturers go to increase
the ability to produce quality audio. It's amazing how little effort and
consideration it takes to marginalize into oblivian a careful studio upgrade
to 24-bit 192KHz.

John


Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Louder IS Better (With Lossy) [email protected] Pro Audio 137 July 13th 03 08:44 AM
Louder IS Better (With Lossy) Richard Kuschel Pro Audio 0 July 1st 03 05:13 AM
Louder IS Better (With Lossy) Chris Johnson Pro Audio 1 July 1st 03 04:22 AM
Louder IS Better (With Lossy) Chris Johnson Pro Audio 0 June 30th 03 11:26 PM
Louder IS Better (With Lossy) Chris Johnson Pro Audio 0 June 30th 03 11:19 PM


All times are GMT +1. The time now is 03:15 PM.

Powered by: vBulletin
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 AudioBanter.com.
The comments are property of their posters.
 

About Us

"It's about Audio and hi-fi"