Lame does not have any specific code to deal with low bitrates. So on 'mid low" (70-100kbps) it can sometimes sound good, but it is likely to fail on some samples. At very low bitrates, it will fail badly in most cases.
FhG encoders have some specific handling for low bitrates.
Guruboolez tried a lot of samples with various MP3 encoders at ~96 kbps here:
http://forum.hardware.fr/hardwarefr/VideoS...jet-84950-1.htm
LAME was clearly the best MP3 encoder and he chose it to be used in the test.
Personally, I have tested WMP10 FhG "CBR 48 kbps", Helix "CBR 48 kbps" and LAME "--preset 48" with a few samples. On average all were equally bad. Only this sample that has extreme stereo separation showed LAME to be clearly worse than the two others:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: twist&shout mp3
Tester: Alex B
1R = E:\test\48kbps\09\MP3Helix - 09 - twist_shout.wav
2L = E:\test\48kbps\MP3FhG - 09 - twist_shout.wav
3R = E:\test\\48kbps\MP3Lame - 09 - twist_shout.wav
---------------------------------------
General Comments: no 3 is clearly the worst. The other two are just very bad.
---------------------------------------
1R File: E:\test\ogg\Test\09\MP3Helix - 09 - twist_shout.wav
1R Rating: 1.5
1R Comment:
---------------------------------------
2L File: E:\test\ogg\Test\09\MP3FhG - 09 - twist_shout.wav
2L Rating: 1.5
2L Comment:
---------------------------------------
3R File: E:\test\ogg\Test\09\MP3Lame - 09 - twist_shout.wav
3R Rating: 1.0
3R Comment:
---------------------------------------
ABX Results:
The used settings were these:
FhG (WMP10): "default 48 kbps CBR"
Helix MP3 (the latest binary from Rarewares): "default 48 kbps CBR" (If I recall correctly I couldn't make Helix to encode 48 kbps VBR)
LAME 3.97b2: --preset 48
The tested samples are from Gabriel's recent 48 kbps test.
The five other samples that I tried didn't show as clear differences:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: chanchanT mp3
Tester: Alex B
1R = E:\test\01\MP3Lame - 01 - chanchanT.wav
2R = E:\test\01\MP3FhG - 01 - chanchanT.wav
3L = E:\test\01\MP3Helix - 01 - chanchanT.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: E:\test\01\MP3Lame - 01 - chanchanT.wav
1R Rating: 1.4
1R Comment:
---------------------------------------
2R File: E:\test\01\MP3FhG - 01 - chanchanT.wav
2R Rating: 1.6
2R Comment:
---------------------------------------
3L File: E:\test\01\MP3Helix - 01 - chanchanT.wav
3L Rating: 1.2
3L Comment:
---------------------------------------
ABX Results:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: fools mp3
Tester: Alex B
1R = E:\test\02\MP3Helix - 02 - fools.wav
2R = E:\test\02\MP3Lame - 02 - fools.wav
3R = E:\test\02\MP3FhG - 02 - fools.wav
---------------------------------------
General Comments: 1 is a bit better with the "breaking glass sound"
1 and 2 are better with voice.
3 is worst
all are bad.
---------------------------------------
1R File: E:\test\02\MP3Helix - 02 - fools.wav
1R Rating: 1.3
1R Comment:
---------------------------------------
2R File: E:\test\02\MP3Lame - 02 - fools.wav
2R Rating: 1.2
2R Comment:
---------------------------------------
3R File: E:\test\02\MP3FhG - 02 - fools.wav
3R Rating: 1.1
3R Comment:
---------------------------------------
ABX Results:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: kraftwerk mp3
Tester: Alex B
1R = E:\test\03\MP3Helix - 03 - kraftwerk.wav
2R = E:\test\03\MP3Lame - 03 - kraftwerk.wav
3R = E:\test\03\MP3FhG - 03 - kraftwerk.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: E:\test\03\MP3Helix - 03 - kraftwerk.wav
1R Rating: 1.4
1R Comment:
---------------------------------------
2R File: E:\test\03\MP3Lame - 03 - kraftwerk.wav
2R Rating: 1.3
2R Comment:
---------------------------------------
3R File: E:\test\03\MP3FhG - 03 - kraftwerk.wav
3R Rating: 1.2
3R Comment:
---------------------------------------
ABX Results:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: sandman mp3
Tester: Alex B
1R = E:\test\06\MP3Helix - 06 - sandman.wav
2R = E:\test\06\MP3FhG - 06 - sandman.wav
3R = E:\test\06\MP3Lame - 06 - sandman.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: E:\test\06\MP3Helix - 06 - sandman.wav
1R Rating: 1.1
1R Comment:
---------------------------------------
2R File: E:\test\06\MP3FhG - 06 - sandman.wav
2R Rating: 1.2
2R Comment:
---------------------------------------
3R File: E:\test\06\MP3Lame - 06 - sandman.wav
3R Rating: 1.0
3R Comment:
---------------------------------------
ABX Results:
ABC/HR for Java, Version 0.5b, 07 june 2006
Testname: Stravinsky_Capriccio mp3
Tester: Alex B
1R = E:\test\07\MP3Lame - 07 - Stravinsky_Capriccio.wav
2R = E:\test\07\MP3Helix - 07 - Stravinsky_Capriccio.wav
3R = E:\test\07\MP3FhG - 07 - Stravinsky_Capriccio.wav
---------------------------------------
General Comments:
---------------------------------------
1R File: E:\test\07\MP3Lame - 07 - Stravinsky_Capriccio.wav
1R Rating: 1.1
1R Comment:
---------------------------------------
2R File: E:\test\07\MP3Helix - 07 - Stravinsky_Capriccio.wav
2R Rating: 1.1
2R Comment:
---------------------------------------
3R File: E:\test\07\MP3FhG - 07 - Stravinsky_Capriccio.wav
3R Rating: 1.1
3R Comment:
---------------------------------------
ABX Results:
I tested these because a web site needed to publish some audio files at about 48 kbps MP3. We ended up using LAME ABR in mono mode because the stereo files were simply too bad.
[!--sizeo:1--][span style=\"font-size:8pt;line-height:100%\"][!--/sizeo--]Edit: typo[/size]