Dear list:
I have been playing recently with lame388b compiled in my
linux machine, trying to experiment with settings to achieve
the best low bit rate possible quality, at the 32 and 24 kbps
mono bitrates. Looking at previous messages regarding this
matter on the ml, i found this very interesting message
(quoted at bottom) from Jaroslav Lukesh about the --highpass
switch, to apply a highpass filter at 54Hz; but when i tried
this with the current beta, the following warning appears:
Warning: highpass filter disabled. highpass frequency too small
So i tried larger values, and found that the smaller value
that lame is accepting, is --highpass 0.3, which gives:
Using polyphase highpass filter, transition band: 0 Hz - 267 Hz
This regardless of the --highpass-width settings, which can
only enlarge, but not reduce the transition band further,
resulting in 300Hz as a minimun for the highpass filter, which,
does not seems very good, in my opinion.
On Thu, 07 Sep 2000, Jaroslav Lukesh wrote:
> | Odesílatel: Gabriel Bouvigne <[EMAIL PROTECTED]>
> |
> | > lame -m m -h -b32 --voice --lowpass 8 --lowpass-width 1
> | > --highpass 0.054 --highpass-width 0.04 --cwlimit 7
> | > --resample 22.05 --noshort a.wav a.mp3
> | >
> | > NOTES:
> | > cwlimit=lowpass minus lowpass-width (kHz)
> |
> | Did you played a lot with cwlimit? Could you please share your results
> about
> | it?
>
> Not as much. But if you use really high bitrates about 160 or 192kbps, then
> you will get better results with higher cwlimit value , such as 11 (by
> default 8.3... kHz).
>
> If you need low bitrate, then set cwlimit as shown, because you do not need
> to calculate tonality over rejected (or partially rejected) frequencies.
>
> If you want to remove highs due low bitrate, then use empiric highpass
> filter set to 0.6/lowpass (kHz), all at -10dB.
> eg., 6kHz lowpass will have 0.6/6=0.1 kHz highpass.
>
> (it is not my empiric, this is from cca 40yrs old book about acoustics)
>
>
> And here LAME values:
>
> --lowpass 6.4 --lowpass-width 1 --highpass 0.06 --highpass-width 0.1
>
> Note that highpass values above 54Hz are very well, because they spare
> first sfb that dataspace should be used for other bands.
>
> -3dB point is at highpass munis highpass-width, or
> lowpass plus lowpass-width
>
> -10dB point you will get at (not measured,
> only speculation, dont know what filters in lame works exactly)
>
> highpass minus highpass-width/3
> or
> lowpass plus lowpass-witdh/3
>
> If you dont apply highpass, then encoded output will sound with very boom
> basses and speech is masked by strong bass.
>
> Regards
>
>
> Jaroslav Lukesh
> --------------------------------------------------
> note: (Bill) Gates to Hell!
>
> --
> MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
--
... "In English, Data." - Crusher
--
MP3 ENCODER mailing list archive is at:
http://www.mail-archive.com/mp3encoder%40minnie.cs.adfa.edu.au/
