Hydrogenaudio Forums > Lame 3.98 beta 8, 2008-04-13

jaybeee

Apr 14 2008, 12:21

Updated : lame 3.98 beta 8 (major new feature) (2008-04-13 15:26)

Sources : http://sourceforge.net/project/showfiles.p...;package_id=309

Binaries : RareWares

LameDropXPd v2 : RareWares LameDrop

CVS : http://lame.cvs.sourceforge.net/lame/lame/...me/?sortby=date

Official LAME v3.98 Beta 8 is out on April 13 2008:

Robert Hegemann:
- LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678
- Found and fixed some suspicious code in additive masking calculation for VBR-NEW
- bug-fix: experimental code was defaulted by accident for VBR-NEW
- fix for some endianess problem on big-endian machines

Changelog - history.html

EDIT: RareWares binaries now available and link to LameDrop added

bug80

Apr 14 2008, 12:30

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

kjoonlee

Apr 14 2008, 13:40

QUOTE (bug80 @ Apr 14 2008, 12:30)

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

Now would be a good time to check if -V5,678 works correctly.

(Some countries use the comma as the decimal thingy.)

shadowking

Apr 14 2008, 13:40

Awesome.. I'll have to try sometime.

Bodhi

Apr 14 2008, 14:31

QUOTE (bug80 @ Apr 14 2008, 13:30)

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

Well, I found it difficult to choose between V2 and V3, so...

sundance

Apr 14 2008, 14:39

QUOTE

Well, I found it difficult to choose between V2 and V3, so...

Me, too. Can't help but thinking of kinda belated April fool's joke...

bug80

Apr 14 2008, 15:09

QUOTE (sundance @ Apr 14 2008, 15:39)

QUOTE

Well, I found it difficult to choose between V2 and V3, so...

Me, too. Can't help but thinking of kinda belated April fool's joke...

I think it's nice. I am looking for a bitrate setting for portable use (as low as possible, without annoying artifacts). I did a lot of ABX testing:

-V5: Annoying artifacts (~130 kbss)
-V4: Not so much annoying artifacts (~165 kbs)

Imagine that "-V4.5" would average around ~148 kbs, without annoying artifacts. On a 8 GB MP3 player, this means almost 800 minutes more music compared to V4!

/mnt

Apr 14 2008, 15:33

Wow this new feature will be very useful and -V2.5 could be enough for anyone. I can not get to compile on Ubuntu to test it though, kept getting a gcc error on the configure script, even if it is installed.

Slipstreem

Apr 14 2008, 15:54

I'd be very interested to know whether the low-pass filter scales linearly with the infinitely variable -V figure. It would offer the interesting possibility of performing an automatic spectrographic scan of any original lossless music content prior to encoding to ensure that we capture everything in terms of upper frequency content in the source material. This would make it possible to have automatic or recommended -V value selection as a new option in front-end GUIs.

Cheers, Slipstreem.

robert

Apr 14 2008, 16:08

QUOTE (kjoonlee @ Apr 14 2008, 14:40)

Now would be a good time to check if -V5,678 works correctly.

No, it does not. It never worked for lowpass too and nobody complaint about it.

QUOTE (bug80 @ Apr 14 2008, 16:09)

I did a lot of ABX testing:

-V5: Annoying artifacts (~130 kbss)
-V4: Not so much annoying artifacts (~165 kbs)

Did you do alot ABX tests with 3.98 b8?

QUOTE (/mnt @ Apr 14 2008, 16:33)

Wow this new feature will be very useful and -V2.5 could be enough for anyone. I can not get to compile on Ubuntu to test it though, kept getting a gcc error on the configure script, even if it is installed.

What's the problem on Ubuntu? Maybe post your problem on lame-dev mailing list and Alexander might help.
Does the following bug-tracker entry match your problem?
https://sourceforge.net/tracker/index.php?f...amp;atid=100290

bug80

Apr 14 2008, 16:24

QUOTE (robert @ Apr 14 2008, 17:08)

QUOTE (bug80 @ Apr 14 2008, 16:09)

I did a lot of ABX testing:

-V5: Annoying artifacts (~130 kbss)
-V4: Not so much annoying artifacts (~165 kbs)

Did you do alot ABX tests with 3.98 b8?

No, v3.97. This was hypothetical, of course I have to redo the tests for 3.98b8.

/mnt

Apr 14 2008, 16:42

QUOTE (robert @ Apr 14 2008, 16:08)

What's the problem on Ubuntu? Maybe post your problem on lame-dev mailing list and Alexander might help.

Oops never mind, I needed to install G++ for it to work. Anyway I just tried at -V 0.8

CODE

lame -V 0.8 Stigmata.wav
LAME 3.98 (beta 8, Apr 14 2008) 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding Stigmata.wav to Stigmata.wav.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=0.8)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
13206/13206 (100%)|    0:14/    0:14|    0:14/    0:14|   23.388x|    0:00
32 [   48] *
40 [    0]
48 [    1] %
56 [    0]
64 [    0]
80 [    0]
96 [    0]
112 [    1] *
128 [    4] %
160 [  224] %**
192 [ 5114] %%%%%%%******************************************************
224 [ 5700] %%%%%%%%%%*********************************************************
256 [ 1386] %%%%%************
320 [  728] %%%%%****
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  218.4       16.3  83.7        94.4   3.2   2.4
Writing LAME Tag...done
ReplayGain: -5.4dB

-V 0

CODE

lame -V 0 Stigmata.wav
LAME 3.98 (beta 8, Apr 14 2008) 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz
Encoding Stigmata.wav to Stigmata.wav.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=0)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
13206/13206 (100%)|    0:15/    0:15|    0:15/    0:15|   22.816x|    0:00
32 [   48] *
40 [    0]
48 [    0]
56 [    0]
64 [    0]
80 [    1] %
96 [    0]
112 [    0]
128 [    1] %
160 [   11] %
192 [  393] %%%%**
224 [ 3757] %%%%%%%%%%%%%%%%%%**********************************
256 [ 4924] %%%%%%%%%%%%%%%%%%%%%%%%*******************************************
320 [ 4071] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*****************
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  263.8       46.6  53.4        94.4   3.2   2.4
Writing LAME Tag...done
ReplayGain: -5.4dB

-V 2.5

CODE

lame -V 2.5 Stigmata.wav
LAME 3.98 (beta 8, Apr 14 2008) 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 17960 Hz - 18494 Hz
Encoding Stigmata.wav to Stigmata.wav.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=2.5)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
13206/13206 (100%)|    0:14/    0:14|    0:14/    0:14|   24.023x|    0:00
32 [   49] %
40 [    0]
48 [    0]
56 [    0]
64 [    0]
80 [    1] *
96 [    0]
112 [   13] *
128 [  244] ***
160 [ 5434] %%%************************************************************
192 [ 5826] %%%%%%*************************************************************
224 [ 1058] %%***********
256 [  337] %%**
320 [  244] %%*
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  183.5        8.0  92.0        94.4   3.2   2.4
Writing LAME Tag...done
ReplayGain: -5.4dB

-V 2

CODE

lame -V 2 Stigmata.wav
LAME 3.98 (beta 8, Apr 14 2008) 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding Stigmata.wav to Stigmata.wav.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=2)
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
13206/13206 (100%)|    0:14/    0:14|    0:14/    0:14|   23.824x|    0:00
32 [   49] %
40 [    0]
48 [    0]
56 [    0]
64 [    0]
80 [    1] *
96 [    0]
112 [    3] *
128 [   51] *
160 [ 2847] %%**************************
192 [ 7053] %%%%%**************************************************************
224 [ 2326] %%%********************
256 [  521] %%%**
320 [  355] %%**
-------------------------------------------------------------------------------
   kbps        LR    MS  %     long switch short %
  195.8       10.2  89.8        94.4   3.2   2.4
Writing LAME Tag...done
ReplayGain: -5.4dB

capma

Apr 14 2008, 16:47

LAME 3.98b8 compiled with ICL 10.1

Download

Silversight

Apr 14 2008, 19:01

Now it's up to the foobar2000 devs to correct the codec profile field. A setting of -V 1.99999 for example results in foobar showing VBR V1.

bhoar

Apr 14 2008, 19:30

QUOTE (/mnt @ Apr 14 2008, 10:42)

Oops never mind, I needed to install G++ for it to work. Anyway I just tried at -V 0.8

CODE

lame -V 0.8 Stigmata.wav...

I take it that's Ministry and you're using it as a worse case scenario (source material contains a lot of distortion and electronically generated sounds/noise)?

-brendan

/mnt

Apr 14 2008, 20:00

QUOTE (bhoar @ Apr 14 2008, 19:30)

QUOTE (/mnt @ Apr 14 2008, 10:42)

Oops never mind, I needed to install G++ for it to work. Anyway I just tried at -V 0.8

CODE

lame -V 0.8 Stigmata.wav...

I take it that's Ministry and you're using it as a worse case scenario (source material contains a lot of distortion and electronically generated sounds/noise)?

-brendan

Yep thats Ministry, some tracks from them can sometimes struggle at -V2. This track has some pre-echo problems on drums at around 2:51 (both 3.90.3 and 3.97) and 3:41 (very easy to ABX on 3.97) at -V2 --vbr-new or preset standard (3.90.3), depending on what sound equipment is used. I could pick it up on my old headphones (sennheiser hd202) but not on my new headphones (sennheiser hd215).

Ojay

Apr 14 2008, 20:17

QUOTE (jaybeee @ Apr 14 2008, 13:21)

bug-fix: experimental code was defaulted by accident for VBR-NEW

Good to see that the new spreading function has been disabled again.

lvqcl

Apr 14 2008, 22:21

Yet another plot (bitrate vs. V value):

There are some jumps on the curve, for example: 2.99 -> 3 (-Y switch?); 6.248 -> 6.249 (resampling to 32 kHz starts here).

(Of course exact bitrate values are valid only for my test wav file. YMMV.

)

kjoonlee

Apr 14 2008, 23:04

QUOTE (robert @ Apr 14 2008, 16:08)

QUOTE (kjoonlee @ Apr 14 2008, 14:40)

Now would be a good time to check if -V5,678 works correctly.

No, it does not. It never worked for lowpass too and nobody complaint about it.

Oh well, I just thought I'd mention it since flac.exe had a bug with that feature once..

Maurits

Apr 14 2008, 23:16

QUOTE (kjoonlee @ Apr 14 2008, 13:40)

QUOTE (bug80 @ Apr 14 2008, 12:30)

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

Now would be a good time to check if -V5,678 works correctly.

(Some countries use the comma as the decimal thingy.)

'Some' countries? The comma is more common than a dot as decimal separator...

Jojo

Apr 15 2008, 06:08

QUOTE (jaybeee @ Apr 14 2008, 04:21)

[*]bug-fix: experimental code was defaulted by accident for VBR-NEW

is that a bug introduced in beta 7 or how long has it been around?

Jebus

Apr 15 2008, 06:31

QUOTE (lvqcl @ Apr 14 2008, 15:21)

Yet another plot (bitrate vs. V value):

There are some jumps on the curve, for example: 2.99 -> 3 (-Y switch?); 6.248 -> 6.249 (resampling to 32 kHz starts here).

(Of course exact bitrate values are valid only for my test wav file. YMMV.

)

Thanks! That's really interesting!

Nick.C

Apr 15 2008, 08:18

QUOTE (Maurits @ Apr 14 2008, 23:16)

QUOTE (kjoonlee @ Apr 14 2008, 13:40)

QUOTE (bug80 @ Apr 14 2008, 12:30)

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

Now would be a good time to check if -V5,678 works correctly.

(Some countries use the comma as the decimal thingy.)

'Some' countries? The comma is more common than a dot as decimal separator...

Looking at the list, India, the Peoples Republic of China, Pakistan, Nigeria and the United States all use the '.' - that's almost half of the population of the globe already....

Squeller

Apr 15 2008, 10:34

QUOTE (Nick.C @ Apr 15 2008, 09:18)

Looking at the list, India, the Peoples Republic of China, Pakistan, Nigeria and the United States all use the '.' - that's almost half of the population of the globe already....

Well, but hmm, big parts of africa ...once again... grey. Unknown territory. Maybe they have extraterrestrial number systems there?

john33

Apr 15 2008, 12:36

Intel 10.1 compiles now at Rarewares.

lamedropXPd has been modified to take account of the new fractional VBR quality settings.

Slipstreem

Apr 15 2008, 13:17

LameDropXPd v2 based on LAME 3.98 beta 8

Cheers, Slipstreem.

willsnotwillis

Apr 15 2008, 15:35

I don't know about anyone else but I think that this version sounds much warmer and "full" than 3.97.
Although I didn't save the results I ABXed them with reasonable ease.

This was with the 3.97 version set to -V 1 and the 3.98b8 version at -V 1.178.
The reason for the strange quality value is that this is the point where the bitrates were the same for the two tracks. This also suggests that the perceived quality of the newer version is better even for lower quality settings.

This might just be my opinion and it is likely that this will change again when a stable version is released.......

DigitalDictator

Apr 15 2008, 16:21

QUOTE (willsnotwillis @ Apr 15 2008, 16:35)

I don't know about anyone else but I think that this version sounds much warmer and "full" than 3.97.
Although I didn't save the results I ABXed them with reasonable ease.

You're kidding, right?

willsnotwillis

Apr 15 2008, 16:25

I'm not kidding but I do admit that it wasn't a very comprehensive test that I carried out!

shadowking

Apr 15 2008, 16:30

'Warmer' at -V1 ??.. Post good abx or get outta here !

willsnotwillis

Apr 15 2008, 16:40

Warmer at V1.178 with the beta.

It is very possible that I am wrong so disagree as much as you want to; if I think that it sounds better on my system with my setup that that is all I want! I have a flac copy of everything anyway so that I can easily change my lossy codec of choice. Tastes change.

Whether or not the beta codec is "better" with reference to the original (ie a lossless copy) is unknown and I accept the outcome either way.

To clarify, I made this conclusion in a short amount of time and I failed to identify the track on a few of the ABX's. Hence the "reasonable ease" clause.

You can tell me to:

QUOTE

Post good abx or get outta here !

if you like but it will not stop me expressing my opinion (unless of course I am blocked from the forums) at the current time.

Alexxander

Apr 15 2008, 17:10

QUOTE (willsnotwillis @ Apr 15 2008, 16:40)

...
You can tell me to:

QUOTE

Post good abx or get outta here !

if you like but it will not stop me expressing my opinion (unless of course I am blocked from the forums) at the current time.

To avoid others taking your opinion as a fact, maybe wrongly, you should backup your statement with ABX test results. You might find out that it's all imagination, or not.

willsnotwillis

Apr 15 2008, 17:21

I completely agree:

QUOTE

To avoid others taking your opinion as a fact, maybe wrongly, you should backup your statement with ABX test results. You might find out that it's all imagination, or not.

and I accept that I was not entirely "correct" in my approach.

When I posted it I did not envisage such an uproar to occur, I merely posted it as a quick and preliminary observation. In the future I realise that ABX results are necessary to back up conclusions or statements.

[JAZ]

Apr 15 2008, 18:12

HydrogenAudio forums are a place for information and knowledge. That is why opinions, especially if they seem weak or improbable, are seen as bad and prevented as much as possible.

Many tests have been done here, and many people have ended realizing that the differences they thought they heard, became just imagination. And i guess you understand that imagination is not knowledge.

So don't take it as people getting onto you for giving your oppinion, but rather to try to keep the rules, which maintain the value of the forums.

zipr

Apr 15 2008, 18:13

Has anyone done any ABX testing of b8 yet?

bhoar

Apr 15 2008, 20:08

QUOTE (willsnotwillis @ Apr 15 2008, 11:21)

When I posted it I did not envisage such an uproar to occur, I merely posted it as a quick and preliminary observation. In the future I realise that ABX results are necessary to back up conclusions or statements.

Then you did not fully read or comprehend the TOS as required to create create your account. If you have not already done so (your last sentence may indicate you have addressed this), reread entry #8 at this link to understand why:

http://www.hydrogenaudio.org/forums/index.php?showtopic=3974

-brendan

greynol

Apr 15 2008, 20:11

The TOS #8 matter has been rectified.

Let's move on.

Ojay

Apr 15 2008, 21:33

QUOTE (bug80 @ Apr 14 2008, 13:30)

QUOTE (jaybeee @ Apr 14 2008, 13:21)

[*]LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like -V5.678

That is a great feature!

Not really useful IMHO except if it is the first step to replace ABR (ABR: quality not much better than CBR)....

lvqcl

Apr 15 2008, 22:36

This feature is useful if you have some target bitrate (various listening tests, for example)

mdefranc

Apr 16 2008, 01:13

The decimal spec also will be useful if your ultimate goal is filling completely a burned cd with a given program.

IgorC

Apr 16 2008, 04:37

Ok, right now here is floating point VBR. from 9,999999 to 0.
But there is still a huge variance of bitrate between VBR 0 and CBR 320.

What about new VBR 0 or "-1" that will be as good as CBR 320.
VBR is better than CBR, so less bitrate then insane profile (-b 320) but same quality.

I could abx V0 but not CBR 320 on this sample http://www.hydrogenaudio.org/forums/index....ost&id=3453

Squeller

Apr 16 2008, 05:23

QUOTE (IgorC @ Apr 16 2008, 05:37)

Ok, right now here is floating point VBR. from 9,999999 to 0.
But there is still a huge variance of bitrate between VBR 0 and CBR 320.

http://www.hydrogenaudio.org/forums/index....st&p=363450:

QUOTE (Gabriel @ Feb 11 2006, 09:49)

[...]yes there is a VBR step missing after V0.
I know that. However, we only have a 10points vbr scale, and we made our best into fitting it to to most usefull steps, so this extra high vbr had to be left out.

Just do an 11 points VBR scale

(-1 to 10). Vorbis like but only vice versa. I doubt I'd need it but hey. The encoder should offer anything that's in it's scope. People decide, what they use.

robert

Apr 16 2008, 11:52

QUOTE (IgorC @ Apr 16 2008, 05:37)

I could abx V0 but not CBR 320 on this sample http://www.hydrogenaudio.org/forums/index....ost&id=3453

As this thread is about 3.98 beta 8 and we don't have many ABX results for it, would you like to share yours with us?

Alexxander

Apr 16 2008, 12:43

QUOTE (robert @ Apr 16 2008, 11:52)

QUOTE (IgorC @ Apr 16 2008, 05:37)

I could abx V0 but not CBR 320 on this sample http://www.hydrogenaudio.org/forums/index....ost&id=3453

As this thread is about 3.98 beta 8 and we don't have many ABX results for it, would you like to share yours with us?

This week I have time and I think my ABX skill is good enough to help. Please tell me which samples to ABX and whether only lossless vs beta8 or also older betas or v3.97

Alexxander

Apr 16 2008, 15:03

QUOTE (robert @ Apr 16 2008, 11:52)

QUOTE (IgorC @ Apr 16 2008, 05:37)

I could abx V0 but not CBR 320 on this sample http://www.hydrogenaudio.org/forums/index....ost&id=3453

As this thread is about 3.98 beta 8 and we don't have many ABX results for it, would you like to share yours with us?

I converted the linked sample with lame3.98b8 -V0 (236 kbps average) because higher V values were too easy to ABX, especially when focussed on the start of the sample.

After finding 2 ABX'able points I stopped searching. The first point obviously was the start of the sample and the second place was between 10.0 and 11.0 secs. I decided to concentrate on the beginning as this point was easier to ABX. I noticed a kind of distortion in the stereo echo of guitar snare hitting, first echo is at about 1.4 sec (I suppose it's a guitar but really don't know much about instruments).

My ABX result:

QUOTE

foo_abx 1.3.1 report
foobar2000 v0.9.5.1
2008/04/16 15:39:34

File A: C:\Documents and Settings\Alexxander\Escritorio\spill_the_blood.flac
File B: C:\Documents and Settings\Alexxander\Escritorio\spill_the_blood.mp3

15:39:34 : Test started.
15:39:43 : 01/01 50.0%
15:39:56 : 02/02 25.0%
15:40:12 : 03/03 12.5%
15:40:20 : 04/04 6.3%
15:40:27 : 05/05 3.1%
15:40:34 : 06/06 1.6%
15:40:42 : 07/07 0.8%
15:40:49 : 08/08 0.4%
15:40:56 : 09/09 0.2%
15:41:09 : 10/10 0.1%
15:41:16 : 11/11 0.0%
15:41:32 : 12/12 0.0%
15:41:54 : 13/13 0.0%
15:41:57 : Test finished.

----------
Total: 13/13 (0.0%)

robert

Apr 16 2008, 15:25

QUOTE (Alexxander @ Apr 16 2008, 13:43)

QUOTE (robert @ Apr 16 2008, 11:52)

QUOTE (IgorC @ Apr 16 2008, 05:37)

I could abx V0 but not CBR 320 on this sample http://www.hydrogenaudio.org/forums/index....ost&id=3453

As this thread is about 3.98 beta 8 and we don't have many ABX results for it, would you like to share yours with us?

This week I have time and I think my ABX skill is good enough to help. Please tell me which samples to ABX and whether only lossless vs beta8 or also older betas or v3.97

A comparison against 3.97 --vbr-new (test for regressions) and 3.98 ABR/CBR would be helpful.

QUOTE (Alexxander @ Apr 16 2008, 16:03)

I converted the linked sample with lame3.98b8 -V0 (236 kbps average) because higher V values were too easy to ABX, especially when focussed on the start of the sample.

After finding 2 ABX'able points I stopped searching. The first point obviously was the start of the sample and the second place was between 10.0 and 11.0 secs. I decided to concentrate on the beginning as this point was easier to ABX. I noticed a kind of distortion in the stereo echo of guitar snare hitting, first echo is at about 1.4 sec (I suppose it's a guitar but really don't know much about instruments).

Thank you Alexxander. Are these distortions in ABR/CBR too? Is there some bitrate sweet-spot where these distortions happen in ABR/CBR too?

Alexxander

Apr 16 2008, 15:47

Encoded spill_the_blood.flac with lame v3.97 V0 (average 222 kbps) and again focussed on first 2.5 secs:

QUOTE

foo_abx 1.3.1 report
foobar2000 v0.9.5.1
2008/04/16 16:28:28

File A: C:\Documents and Settings\Alexxander\Escritorio\spill_the_blood\spill_the_blood.flac
File B: C:\Documents and Settings\Alexxander\Escritorio\spill_the_blood\spill_the_blood 397-V0.mp3

16:28:28 : Test started.
16:28:42 : 01/01 50.0%
16:28:48 : 02/02 25.0%
16:28:56 : 03/03 12.5%
16:29:01 : 04/04 6.3%
16:29:07 : 05/05 3.1%
16:29:13 : 06/06 1.6%
16:29:18 : 07/07 0.8%
16:29:35 : 08/08 0.4%
16:29:53 : 09/09 0.2%
16:30:10 : 10/10 0.1%
16:30:16 : 11/11 0.0%
16:30:23 : 12/12 0.0%
16:30:30 : 13/13 0.0%
16:30:48 : 14/14 0.0%
16:30:51 : Test finished.

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

As far as this sample concerns v3.98b8 sounds better than v3.97 (both V0 vbr-new). v3.97 was easier to ABX, I had no doubts matching. v3.98 V0 required good concentration, hearing fatigue played part.

Tomorrow I will ABX CBR 320 kbps but what bitrate do you propose for ABR? (I never have used ABR and for the last 5 years only used VBR).

Edit: I saw robert posted while I was writing this post and answered my doubt about ABR. I will try to find the ABR bitrate sweet-spot.

DJED

Apr 16 2008, 17:47

Gang -

Simple me again...

I use V0 (alt preset extreme) pretty much all the time. The other settings are not of use.

Currently, I have an earlier 3.98 build. What can I expect in w/ this release?

Goal = MAX VBR flexibility while having closest to transparent as possible. I am not a fan of CBR.

Is V0 still my best option should I use this release?

Thanks gang!

ZinCh

Apr 16 2008, 18:09

Where I can read more about:

CODE

experimental switches:
-Y lets LAME ignore noise in sfb21, like in CBR

Alexxander

Apr 17 2008, 08:58

QUOTE (robert @ Apr 16 2008, 15:25)

Thank you Alexxander. Are these distortions in ABR/CBR too? Is there some bitrate sweet-spot where these distortions happen in ABR/CBR too?

I have done more ABX-ing. I could not ABX v3.97 CBR 320 kbps so I played with ABR v3.97 and 3.98b8. Converted tracks were easily identified when encoded with --ABR 200 and --ABR 230 so I ABX-ed --ABR 250. This was as hard for v3.97 (243 kbps) as for 3.98b8 (242 kbps) but ABX-able without guessing. Then I encoded with --ABR 270 (both resulted in 255 kbps), at this ABR speed ABX-ing became rather impossible and I started to guess. I focussed always on the same distortion at the beginning of the sample (described in my posts above) and could not tell whether v3.97 sounded better or worse than v3.98b8 (for --ABR 250). Refining will be hard but if it's usefull I will give it a try. Maybe ABX-ing an other sample would be more helpful.