IPB

Welcome Guest ( Log In | Register )

2 Pages V   1 2 >  
Reply to this topicStart new topic
LAME mp3 cbr 320 vs vbr -V0, Which one has better quality
eboyer93
post Dec 9 2008, 22:00
Post #1





Group: Banned
Posts: 113
Joined: 4-November 07
Member No.: 48453



Which one has better quality?
Go to the top of the page
+Quote Post
pdq
post Dec 9 2008, 22:09
Post #2





Group: Members
Posts: 3426
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



One would tend to guess cbr 320, but you would be hard-pressed to hear any difference in most material.
Go to the top of the page
+Quote Post
greynol
post Dec 9 2008, 22:19
Post #3





Group: Super Moderator
Posts: 10092
Joined: 1-April 04
From: San Francisco
Member No.: 13167



...and if anyone can, please present your blind test results here(*).

There is no difference in quality for those who cannot.

(*) Also, please consider weighing in on this discussion:
http://www.hydrogenaudio.org/forums/index....showtopic=67840

This post has been edited by greynol: Dec 9 2008, 22:20


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
eboyer93
post Dec 9 2008, 22:27
Post #4





Group: Banned
Posts: 113
Joined: 4-November 07
Member No.: 48453



Thanks
Go to the top of the page
+Quote Post
xSerpentx
post Dec 13 2008, 03:00
Post #5





Group: Members
Posts: 4
Joined: 5-March 07
Member No.: 41190



-V0 is better in many ways.


--------------------
EAC | LAME3.98 -V4
Go to the top of the page
+Quote Post
Synthetic Soul
post Dec 13 2008, 09:52
Post #6





Group: Super Moderator
Posts: 4887
Joined: 12-August 04
From: Exeter, UK
Member No.: 16217



QUOTE (xSerpentx @ Dec 13 2008, 02:00) *
-V0 is better in many ways.
  1. Smaller filesize.
Ummm...


--------------------
I'm on a horse.
Go to the top of the page
+Quote Post
Shunsuke_01
post Jan 8 2009, 16:35
Post #7





Group: Members
Posts: 24
Joined: 17-December 07
From: Australia
Member No.: 49623



QUOTE (Synthetic Soul @ Dec 13 2008, 19:52) *
QUOTE (xSerpentx @ Dec 13 2008, 02:00) *
-V0 is better in many ways.
  1. Smaller filesize.
Ummm...

I think that's about it. laugh.gif
Go to the top of the page
+Quote Post
Blai
post Jul 29 2009, 17:01
Post #8





Group: Members
Posts: 5
Joined: 29-July 09
From: Barcelona, Spain
Member No.: 71879



So I end up assuming that CBR 320 kbps and VBR -V0 have the same quality and the only difference is file size.
Go to the top of the page
+Quote Post
pdq
post Jul 29 2009, 17:27
Post #9





Group: Members
Posts: 3426
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



Except for some extremely rare examples, yes.
Go to the top of the page
+Quote Post
kawaiigardiner
post Aug 3 2009, 08:52
Post #10





Group: Members
Posts: 32
Joined: 3-August 09
Member No.: 71985



Having had a look at the relevant wiki entry on hydrogenaudio - it appears that 320 has an insignificant improvement but the amount of space for that improvement really can't be justified. I've got around 100GB of music using -V0 using 3.98.2 and I can't tell the difference.
Go to the top of the page
+Quote Post
d_headshot
post Aug 6 2009, 18:55
Post #11





Group: Members
Posts: 195
Joined: 28-September 08
Member No.: 58729



QUOTE (Shunsuke_01 @ Jan 8 2009, 10:35) *
QUOTE (Synthetic Soul @ Dec 13 2008, 19:52) *
QUOTE (xSerpentx @ Dec 13 2008, 02:00) *
-V0 is better in many ways.
  1. Smaller filesize.
Ummm...

I think that's about it. laugh.gif


Doesn't it have a better encoding algorithm as well?
Go to the top of the page
+Quote Post
pdq
post Aug 6 2009, 20:22
Post #12





Group: Members
Posts: 3426
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



QUOTE (d_headshot @ Aug 6 2009, 13:55) *
Doesn't it have a better encoding algorithm as well?

Better in what sense? If you mean that it achieves better sound quality, then no.
Go to the top of the page
+Quote Post
Destroid
post Aug 7 2009, 01:44
Post #13





Group: Members
Posts: 555
Joined: 4-June 02
Member No.: 2220



Has anyone dared to estimate the scale for law of diminishing returns for LAME? It might help with some of these posts that ask, "Is X setting better than Y?"

example:
-V 5 = 95% likely to be transparent
...
-V 2 -Y = 98% transparent
-V 2 = 98.5 % transparent
...
-V 0 = 99% transparent
-b 320 = 99.2% transparent

Of course, it's not supposed to be empirical, just some stats based on other members' posts.

The point is, that the bit rate trade-off is less justified.


--------------------
"Something bothering you, Mister Spock?"
Go to the top of the page
+Quote Post
psycho
post Aug 7 2009, 08:30
Post #14





Group: Members
Posts: 241
Joined: 14-October 05
Member No.: 25099



I'd like to add something in layman's words... Even though VBR mode (0) doesn't use 320 kbps for every frame, as offcourse CBR 320 mode does, it has to my understanding a more sophisticated way of using psychoacoustics and such so it can actually happen that the CBR 320 mode has some artifacts, whereas VBR mode 0 does not! So, the best VBR mode > any CBR mode! That goes even for ABR, so I dare to say it's actually like this: VBR > ABR > CBR in terms of probable quality one gets. Every mode offcourse has it's flaws, but the VBR algorithm has been worked on to oblivion, whereas the CBR is obsolete these days, IMHO.


--------------------
lame -V 0
Go to the top of the page
+Quote Post
Synthetic Soul
post Aug 7 2009, 09:08
Post #15





Group: Super Moderator
Posts: 4887
Joined: 12-August 04
From: Exeter, UK
Member No.: 16217



QUOTE (Destroid @ Aug 7 2009, 01:44) *
Has anyone dared to estimate the scale for law of diminishing returns for LAME? It might help with some of these posts that ask, "Is X setting better than Y?"
There is the graph in the wiki, but it meant only as a very general guide (quality scores provided by Gabriel with file sizes for one album ).

QUOTE (psycho @ Aug 7 2009, 08:30) *
I'd like to add something in layman's words... Even though VBR mode (0) doesn't use 320 kbps for every frame, as offcourse CBR 320 mode does, it has to my understanding a more sophisticated way of using psychoacoustics and such so it can actually happen that the CBR 320 mode has some artifacts, whereas VBR mode 0 does not!
This is the second suggestion that -V0 has a better algorithm. Can anyone accurately deny or confirm this?


--------------------
I'm on a horse.
Go to the top of the page
+Quote Post
shadowking
post Aug 7 2009, 09:28
Post #16





Group: Members
Posts: 1527
Joined: 31-January 04
Member No.: 11664



V0 is never better. The psy parameters are not worse either and actually more conservative. You can check with --verbose.



--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
[JAZ]
post Aug 7 2009, 10:12
Post #17





Group: Members
Posts: 1793
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



It is true that the analysis done while encoding in VBR is different than the one done for CBR.

From the help that I updated to be included in an upcoming version of LAME: (if they still remember... wink.gif )
QUOTE
The difference between ABR and true VBR is in how the desired number of bits is chosen. The true VBR mode determines the number of bits based on the quantization noise. VBR figures out how many bits are needed so that the quantization noise is less than the allowed masking.

ABR mode uses the CBR formula to determine the desired number of bits. This formula is based on the perceptual entropy, which is a rough measure of how difficult the frame is to encode.

(Note that this also denies that ABR is the same than VBR)

So yes, there exist examples where an VBR encode does not show an artifact that a CBR 320 encode does ( and viceversa ).
Go to the top of the page
+Quote Post
shadowking
post Aug 7 2009, 13:39
Post #18





Group: Members
Posts: 1527
Joined: 31-January 04
Member No.: 11664



-V0 is targeting ~ 224k . It could be competitive with 256k cbr?? but I've never seen it beat 320k or --abr 287 -h. There are still some problem samples for vbr and maybe vice versa. In abx tests i've done --abr 287 was always better when V0 had artifacts.

In short it may be that ABR suffers from a different artifact, Yet i'm not convinced that VBR 224k can take on 270 ~ 320k ABR. It might have chance if the psymodel was perfect but its not and there is too much bitrate difference.


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
d_headshot
post Aug 7 2009, 20:47
Post #19





Group: Members
Posts: 195
Joined: 28-September 08
Member No.: 58729



QUOTE (shadowking @ Aug 7 2009, 07:39) *
-V0 is targeting ~ 224k . It could be competitive with 256k cbr??


I don't think a V0 could compete against 256 kbps in that type of comparison. The ABR of a VBR file doesn't determine it's general quality. It's each frame that has to be compared because you may have some frames going as low as 64 kbps if there is silence, or as high as 320 kbps if there is a heavy chorus.
Go to the top of the page
+Quote Post
psycho
post Aug 7 2009, 21:12
Post #20





Group: Members
Posts: 241
Joined: 14-October 05
Member No.: 25099



d_headshot, I think you have a very arguable point here. The fact that -V 0 goes as high as 320 kbps for "complex" frames, suggests that it potentially can cope with those "complex" frames better than any CBR mode lower than 320 kpbs can... With this logic, -V 0 beats any CBR mode, except arguably 320 kbps.

Offcourse, my theory here is based only upon logical thinking and not on tests I have done, unlike shadowking's findings, for which he has test results to support his claims.

edit: Unfortunately we are debating about such high bitrates, that we can not ABX test our theories, as all of these modes are transparent, except for problem samples, which exist for all of them. So, it's a difficult debate. We can only debate indefinitely which mode is potentially better. Or maybe we could count problem samples found for each one of the modes and the one that has fewer of those is better. But that can never be achieved, because we just don't have all the possible sounds that exist to test them all...

I think we are not going to reach an undisputable conclusion.

This post has been edited by psycho: Aug 7 2009, 21:18


--------------------
lame -V 0
Go to the top of the page
+Quote Post
halb27
post Aug 7 2009, 22:25
Post #21





Group: Members
Posts: 2439
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (psycho @ Aug 7 2009, 22:12) *
... The fact that -V 0 goes as high as 320 kbps for "complex" frames, suggests that it potentially can cope with those "complex" frames better than any CBR mode lower than 320 kpbs can... With this logic, -V 0 beats any CBR mode, except arguably 320 kbps. ...

This is a theretical consideration, and as such it is not correct.
Because of the bit reservoir there is no direct relation between frame bitrate and audio data bitrate. CBR 256 can have a very high audio data bitrate at problematic spots because it can use bits from the previous frame. On the other hand the bits available for audio data purposes within a 320 kbps frame can remain unused at critical spots. In fact because of a certain compatibility precaution for mp3 playback with WMP and due to some inaccuracy in the mp3 specs regarding the audio data capacity of 320 kbps frames current LAME doesn't make full use of the audio data capacity of 320 kbps frames. The latter fact is the reason why the lossless mp3packer procedure reduces filesize of CBR 320 or very high bitrate ABR encodings significantly. mp3packer has the same positive effect on the 320 kbps frames of VBR encodings. (BTW no problem is known for mp3packer improved Lame encodings, even for WMP.)

Which doesn't tell about quality, but shows that a simple frame bitrate based theoretical discussion doesn't help.

With a good VBR implementation as done with 3.98, -V0 is expected to usually yield the better quality compared to CBR 256 in those rare cases that there is an audible difference between -V0 and CBR 256. It may not be so in every such situation.
As long as ABR and CBR mode is implemented correctly CBR 320 and very high bitrate ABR (for instance shadowking's ABR 287 -h setting) are expected to usually yield the better quality compared to -V0 in those rare cases that there is an audible difference between -V0 and CBR 320 resp. ABR very high bitrate. It may not be so in every such situation.

There had been some suspicion that CBR and ABR mode aren't implemented fine with 3.98, but there has never been real evidence for this IIRC.

/mnt gave a sample recently where -V0 failed but ABR 287 suuceeded, and shadowking made the same experience.

It always comes to the point where to put the sweet point, and this is a very personal decision.
-V0 is so good that not many people really need a better quality setting.
On the other hand with today's storage technology even on DAPs many people nowadays and more so in the near future don't have to care about file size and can afford the best quality mp3 can offer.
Moreover because of their robust quality headroom CBR 320 and very high bitrate ABR are kind of an insurance against situations where the implemented psy model fails unusually badly. This was the case for 3.97 for the sandpaper noise problem and for all versions prior to 3.98 with the trumpet sample. -V0 was real nasty in these cases because VBR heavily relies on the psy model which failed here, and CBR 320 or very high bitrate ABR improved things a lot. No such failure is known for 3.98, looks like 3.98 -V0 is robust against nasty results.

This post has been edited by halb27: Aug 7 2009, 22:55


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
antman
post Aug 7 2009, 22:56
Post #22





Group: Members
Posts: 103
Joined: 3-January 07
From: Texas
Member No.: 39241



QUOTE (shadowking @ Aug 7 2009, 07:39) *
--abr 287 -h


You've brought this bitrate up before, why 287?
Go to the top of the page
+Quote Post
/mnt
post Aug 8 2009, 00:02
Post #23





Group: Members
Posts: 697
Joined: 22-April 06
Member No.: 29877



Sometimes V0 can sound worse then V2 or CBR, but it rarely happens though.

A couple of examples:

LAME 3.97 -V0 --vbr-new

CODE
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/07 23:33:02

File A: C:\Music\Metallica\Ride The Lightning\01. Fight Fire With Fire.mp3
File B: E:\Music\Albums\Metallica - Ride The Lightning\01. Fight Fire With Fire.flac

23:33:02 : Test started.
23:33:21 : 01/01 50.0%
23:33:27 : 02/02 25.0%
23:33:34 : 03/03 12.5%
23:33:38 : 04/04 6.3%
23:33:46 : 05/05 3.1%
23:33:51 : 06/06 1.6%
23:33:55 : 07/07 0.8%
23:34:01 : 08/08 0.4%
23:34:07 : 09/09 0.2%
23:34:18 : 10/10 0.1%
23:34:29 : 11/11 0.0%
23:34:41 : 12/12 0.0%
23:34:43 : Test finished.

----------
Total: 12/12 (0.0%)


Flanging on the harpicord at 0:23 (yes there is a harpicord on a Metallica song).

V2 vs V0 (LAME 3.97)


CODE
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/07 23:35:22

File A: C:\Music\Metallica\Ride The Lightning\01. Fight Fire With Fire.mp3
File B: C:\Temp\Fight Fire With Fire (LAME 3.97 -V2 --vbr-new).mp3

23:35:22 : Test started.
23:36:06 : 01/01 50.0%
23:36:13 : 02/02 25.0%
23:36:23 : 03/03 12.5%
23:36:30 : 04/04 6.3%
23:36:35 : 05/05 3.1%
23:36:43 : 05/06 10.9%
23:36:50 : 06/07 6.3%
23:36:59 : 07/08 3.5%
23:37:04 : 08/09 2.0%
23:37:19 : 09/10 1.1%
23:37:27 : 10/11 0.6%
23:37:45 : 11/12 0.3%
23:37:52 : 12/13 0.2%
23:37:58 : 13/14 0.1%
23:38:03 : 14/15 0.0%
23:38:06 : Test finished.

----------
Total: 14/15 (0.0%)


Artifact is gone or possibly softer on V2.


LAME 3.97 -V0 --vbr-new

CODE
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/07 23:23:36

File A: C:\Music\Gary Numan\The Pleasure Principle\06. Tracks.mp3
File B: E:\Music\Albums\Gary Numan - The Pleasure Principle\06. Tracks.flac

23:23:36 : Test started.
23:24:38 : 01/01 50.0%
23:24:43 : 02/02 25.0%
23:24:50 : 03/03 12.5%
23:25:07 : 04/04 6.3%
23:25:24 : 05/05 3.1%
23:25:33 : 06/06 1.6%
23:25:40 : 07/07 0.8%
23:26:02 : 08/08 0.4%
23:26:06 : 09/09 0.2%
23:26:11 : 10/10 0.1%
23:26:29 : 11/11 0.0%
23:26:35 : 12/12 0.0%
23:26:39 : Test finished.

----------
Total: 12/12 (0.0%)


Drum smear at 2:07.

V2 vs V0 (LAME 3.97)

CODE
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/07 23:27:01

File A: C:\Music\Gary Numan\The Pleasure Principle\06. Tracks.mp3
File B: C:\Temp\Tracks (LAME 3.97 -V2 --vbr-new).mp3

23:27:01 : Test started.
23:27:30 : 01/01 50.0%
23:27:37 : 02/02 25.0%
23:27:44 : 03/03 12.5%
23:27:48 : 04/04 6.3%
23:27:53 : 05/05 3.1%
23:28:05 : 06/06 1.6%
23:28:12 : 07/07 0.8%
23:28:19 : 08/08 0.4%
23:28:25 : 09/09 0.2%
23:28:31 : 10/10 0.1%
23:28:39 : 11/11 0.0%
23:28:44 : 12/12 0.0%
23:28:59 : Test finished.

----------
Total: 12/12 (0.0%)


No drum smearing on the V2 encode.

It keeps me stable for days
In FLAC

QUOTE (shadowking @ Aug 7 2009, 13:39) *
-V0 is targeting ~ 224k . It could be competitive with 256k cbr?? but I've never seen it beat 320k


I find this sample to sound better at V0 then 320 CBR on LAME 3.98.2, due to it's new psy model that --vbr-new now uses.

CODE
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/07 23:18:44

File A: C:\Temp\Show Me Your Spine Sample (LAME 3.98.2 -V0).mp3
File B: C:\Temp\Show Me Your Spine Sample (LAME 3.98.2 b320).mp3

23:18:44 : Test started.
23:18:55 : 01/01 50.0%
23:19:02 : 02/02 25.0%
23:19:14 : 03/03 12.5%
23:19:20 : 04/04 6.3%
23:19:27 : 05/05 3.1%
23:19:35 : 06/06 1.6%
23:19:43 : 07/07 0.8%
23:19:58 : 08/08 0.4%
23:20:06 : 09/09 0.2%
23:20:21 : 10/10 0.1%
23:20:32 : 11/11 0.0%
23:20:40 : 12/12 0.0%
23:20:45 : Test finished.

----------
Total: 12/12 (0.0%)


IMO V0 sounded better then the 320kbps encode. Although both encodes are far away from being transparent.


--------------------
"I never thought I'd see this much candy in one mission!"
Go to the top of the page
+Quote Post
shadowking
post Aug 8 2009, 03:25
Post #24





Group: Members
Posts: 1527
Joined: 31-January 04
Member No.: 11664



QUOTE (antman @ Aug 8 2009, 07:56) *
QUOTE (shadowking @ Aug 7 2009, 07:39) *
--abr 287 -h


You've brought this bitrate up before, why 287?



This is to get an in-between 256 ~ 320k setting. With lame 3.98 there is a restriction on bit reserve that bloats bitrate at very high bitrate, so -h works a little better and 288 is too close to 320k.


--------------------
Wavpack -b450s0.7
Go to the top of the page
+Quote Post
infernovip
post Aug 9 2009, 18:43
Post #25





Group: Members
Posts: 3
Joined: 9-August 09
Member No.: 72167



plis i'm looking 2 days for some 100% perfect information about that:

i don't care about the mp3 file size its can be 100MB i don't care:]
but plizzz tell me
how to make the best sound quality of the mp3

for examlpe
if im using easy cd-da extractor 12.0.1 build 1
you can download fully functional trial http://www.poikosoft.com/

and i want to make the best quality mp3 the best of the best
from my orginal cd audio
i should use

stereo or joint stereo?
44100 or 48000
cbr or vbr?
if cbr then 320
if vbr???
new vbr
compresion lvl0

???
plizzzz someone who know enything about this ripping to mp3 for a anserw

This post has been edited by infernovip: Aug 9 2009, 19:13
Go to the top of the page
+Quote Post

2 Pages V   1 2 >
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 1st November 2014 - 12:07