i want to rip cd with lame.
need to save all bandwidth below 20000HZ.
i try the "-V0 --lowpass 20000",but no luck it just cut at 19000HZ.

i can not figure it out,some one help me please. Thank you so much!

just a question:
do you need to retain frequencies up to 20 kHz for testing purposes

or do you think you need them becaus you think you can hear them?

I have no idea why the lowpass doesn't work, it should. Maybe raising the lowpass a bit higher will help? Also, have you checked that your source really contains such high frequencies?

The other question is, why would you ever want to do that?

Edit> there is a 20khz limitation in the psychoacoustic model in Lame, so you will never get frequencies above that.

high CBR bitrates usually encode more high frequencies than VBR.
Don't know why

yes you are right,test purpose .(seems some mp3s in my download folder have all the frequencies below 20kHz)

absolutely the sourse contains up to 21000Hz.
like what i said, "lowpass 20000" parameter will only up to 19000Hz.
I have tried these command line in foobar2000 diskwriter below:
-k -V 0 --vbr-new -m s --athshort --athtype 2 - %d
--lowpass 20000 -V 0 --vbr-new -m s --athshort --athtype 2 - %d
-b 320 --lowpass 20000 -m s
--alt-preset insane - %d
lame version is 3.96.1
3.97 and 3.98 just give a worse encode than 3.96 with these parameters(i examine them in audition to see visualized).

not at the highest frequencies

jako, something is clearly hosed on your end. I just tested various --lowpass settings with Lame 3.97 and they worked fine. I did all these tests at the command line and I could clearly see that, during the encode, the "polyphase lowpass filter" changed depending on what I set. Then, after encoding, I let encspot check the files and that confirmed that the lowpass frequency had been changed. I also tried the -k setting and that worked too. Again, I did all this at the command line, and it worked just fine.

dear gib,have you check them in adobe audition(cooledit)?

at "lowpass 19000" or below it works for me,but not when higher than this.

No, but why do I need to? During the encode Lame told me it was using the correct lowpass. After encoding, encspot confirmed that. Are you saying that both Lame and encspot are wrong?

But, that being said, I went ahead and did a quick test for the hell of it. I don't own CoolEdit or any commercial software like that, so I used Audacity. In Audacity I created a 5 second clip of white noise, since white noise obviously contains information up and over 20000 Hz. I then encoded that short wav with Lame 3.97 using three different commands: -V0 (for the default), -V0 --lowpass 16 (to use a reduced lowpass), and -V0 -k (to not use any filters at all). A quick look at the frequency analysis in Audacity showed that the different mp3s indeed had different lowpass cutoffs, with the -V0 -k encode going right up to about 20000 Hz before dropping off. It all worked perfectly, as expected.

I think the problem here might be that you're looking at a frequency analysis of music that simply doesn't have much of any content over 19KHz, so the graph looks the same regardless of a 19KHz or 20KHz lowpass filter.

The other problem with using MP3 to try to produce frequency spectra that match the original WAV is that any decent encoder should be trying to throw away the information that is least audible according to its psychoacoustic model. More precisely it will encode the least audible frequencies with lower precision, allowing more noise/distortion so long as it is below the masking level. Sometimes this will result in completely omitting certain frequencies.

Thank you so kind to test.
i tried too like you said,the noise can upto 20000Hz,but my music can't.
my music file's frequency is absolutely up to higher than 20000Hz.
but seems not that strong as the nosie at high frequency(just conclude from the spectrum in audition).

The 19khz dropoff is an ATH 'issue'* with lame 3.94 and higher. Use --noath to get rid of that 'problem'*.

Lame 3.93.1 and earlier do not have the 19khz dropoff.
(this is actually one of the reasons I use lame3.93.1)

Um, not that my tests have shown. I've seen plenty of noise ~21khz (using analfreq 1.8).


*may or may not be an issue/problem, it just 'is'

yes ,thanks!

'analfreq'

Sorry