Daniel,

In CBR mode, LAME in some ways ignores frequencies above
16 kHz, so not using a low-pass filter may allow more
distorted signals at those higher frequencies (as opposed
to mostly blocking the signals at those frequencies).

Considering this, and that only strange and very unusual
experimental audio has audible frequency content nearly all
the way to 22 kHz (the maximum storable frequency at the CD 
sampling rate), try guessing what removing the low pass
will actually do.  Many times, there won't be audible 
content over 19 kHz, so in these cases, the unfiltered
sample will sound exactly the same as a sample with a
low-pass at 19 kHz.  (Okay, practically, that may not be 
true.  The equipment will still try to produce frequencies 
that are inaudible, which eats power, and, depending on the
equipment, could lead to distortion effects that trickle
down into the audible frequency spectrum.)

With VBR, LAME pays more attention to frequencies above
16 kHz, but due to a limitation in the MPEG layer-3
format, accurately preserving those frequencies may cost 
many more bits than it should.  If your motivation for
using "-k" is to attempt to preserve higher frequencies,
you may find better performance with one of the VBR
profiles than with CBR.  However, at a given target bit
rate, the excessive amount of bits sometimes required to
preserve high frequencies may be better spent on the low
frequencies instead.  Many of LAME's VBR profiles already
take this factor into account, and purposefully choose to
partially ignore some of the high frequencies, and instead
try to spend bits where there is a greater overall
reduction in audible distortion per bit.

You might find reason to use "-k" if the audio to be
encoded has already been fed through a low-pass filter, 
or if you are dealing with lower sample rates, like 24 kHz,
where the maximum reproducible frequency is quite audible.

Here are some other questions that you might or might not
care to look into, or that others might comment upon:

 With LAME, does "-q 1" offer better performance than
 "-q 0"?

 Is distrusting the joint stereo code and falling back
 to "-m s" worthwhile, considering the higher bit rate
 needed or possible loss of quality in other areas?

 At 320 kbps, if "-m s" is omitted, is spending the
 maximum bits per second still worth the incremental gain
 in quality, after considering the practical limitations
 of the encoder and MPEG layer-3 audio?

Kind regards,

- John


Johnny Bravo wrote:
> I am new to MP3 encoding.
> 
> 1. What advantages / disadvantages does the command
> "LAME -b 320 -q 0 -m s -k" have over "LAME -b 320 -q 0
> -m s".
> 
> 2. According to the LAME version 3.91 MMX
> (http://www.mp3dev.org/) manual
> 
> -k   keep all frequencies.  (Disable all filters)
> 
> -k will disable all lowpass filtering.  Not
> recommended.
> 
> Does the "Not recommended" statement apply to all CBR
> bitrates (including 256 and 320)?
> 
> Many Thanks
> Daniel X
_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder

Reply via email to