Hi!

I have tested some tracks with Lame 3.97 with "-V2 --vbr-new" and "lame.exe -V2 --vbr-new --lowpass 16000" options. I thought i will have smaller bitrate and at least the same quality with lowpass filter at 16Khz (-V2 dosn't have -Y key inside right?), but that sample with lowpass filter sounds worse. It sounds like in case of lowpass there are bigger pre-echo and attacks more "washed". I tested that sample with ABX test and have following results:

Original track VS "lame.exe -V 2 --vbr-new"


Original track VS "lame.exe -V 2 --vbr-new --lowpass 16000"


"lame.exe -V 2 --vbr-new" VS "lame.exe -V 2 --vbr-new --lowpass 16000"


Original track at .flac format with mp3s encoded. There are album replaygain applied at flac and mp3gain at mp3s: Madonna - Music - Paradise (not for me) (about 59 Mb) Moderation: Link removed - please keep clips under 30s (see TOS #9).

Original track cut at .flac format with mp3s encoded (first 13 seconds): Madonna - Music - Paradise (not for me) cut (about 1.4 Mb)

You can hear that rigth at first five seconds of the track.

I think it is not normal that sample sounds worse with "--lowpass 16000" command prompt key, isn't it?

Thank You.

edit: code to codebox, some typo and gramatics

You say that you applied mp3gain to the mp3s. Are you sure that they are of equal loudness? And are you sure they are both equally loud as the original?

There is 0.02 dB difference - it is inaudible.

Of course i have "Use ReplayGain" option at foobar - preferences - ABX comparator is ON.

When performing ABX tests, you shouldn't use any sort of gain, since the process of encoding can affect the percieved loudness of the track to the replaygain algorithm.
Simply try ABXing the tracks with no replaygain applied.

That said, it is not surprising that the lowpassed encoding sounded worse, since the -Vx presets are tuned for optimal quality without having to resort to other command line switches.

It is normal, otherwise the setting would have been part of the preset.

Instead of using --lowpass, try using -Y instead, and see if you have the same result.

Ok. I'll try that way. But tag says that there is only 0.02 dB difference between that tracks. And the reason why i started the comparison is at i guess there is more noise or preecho not louder.



It is JUST a lowpass filter it is not a tweak to lame internal processes behaviour.


-Y switch change algorithms behaviour but --lowpass is just LPF and i think it should be better to use LPF at 16 Khz than -Y switch. Lowpass shouldn't change lame psymodel behaviour.
But the track starting to sounds worse and it is discouraged for me, why?

A localised energy surge in time domain implies a broad frequency distribution in the frequency domain. Thus it is possible that a lowpass might decrease "localisation" of the surge in the time domain.

You should not forget that time domain and frequency domain are just one transform away from each other. They are two representations of the same data, so if you act on one, you are also acting on the other.

Wow, I like that dirty talk!!!

-V 2 --vbr-new: lowpass = ~18khz
-V 2 --vbr-new --lowpass 16000: lowpass = ~ 16000khz, that effectively shuts OFF the lowpass filter. (unless it's smart enough to know it's wrong..)
Using "--lowpass 16" will give you a 16khz lowpass. (I hope this is what you really used)

As long as you can hear above 16khz, you SHOULD be able to hear the difference.


Using -Y or --lowpass 16 with -V 2 should result is about the same average bitrate, but they probably won't sound the same (depends on the track). Neither of them 'change the psymodel behavior' as far as I can tell, but they DO effect the frame size selection, and thus, the average bitrate.

Lame.exe should be smart enough to understand 16kHz in this case, if I remember well.

Being able to hear a tone above 16kHz is no guarantee that you can hear the difference between a 16 kHz lowpass vs. no lowpass when listening to actual music. I recall a test here some time back that showed that most of us cannot hear a 16 kHz lowpass even in music with significant energy above this frequency.

I agree, I can hear a tone up to 22 khz but I failed to abx files with 16khz lowpass and up in the second test :

Test cutoff
Test lowpass

It's smart enough.

I guess i don't like the -Y hi frequency fluctuations, i thing that simple cut off all above 16Khz is a better way.
Is there any conclusion which method is better?

My hearing frequency limit is about 17500 Hz.

Cool.


True (for me as well), but it also depends on the music itself.


I would tend to agree. I find -V 3 & -V 4 to be occasionally 'chirpy' with stuff above 16khz, but I don't think it is due to -Y only (ATH maybe?).