Skip to main content
Topic: LAME applying 16kHz cutoff at 320K CBR (Read 19729 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

LAME applying 16kHz cutoff at 320K CBR

Reply #25
Above 16KHz is a very small part of the highest audible octave. Most people probably cant even hear it. In many cases, filtering out these frequencies will make many recordings sound better, even if not accurate... if you even notice..

Anyway, FLAC ftw.

LAME applying 16kHz cutoff at 320K CBR

Reply #26
@pdq I agree with your post, especially after the responses from [JAZ] and Dynamic

@greynol LAME is expected to filter out frequencies above 20.5kHz @ 320kbps since it's default lowpass is set at 20.5, and to an extent keep frequencies up to 20.5kHz.  Shouldn't it?  (Is that what was called begging the question?)

I guess you're trying to throw the flak with your responses by answering a question with another question, when in fact, an inquirer expects an declarative answer rather than with another question.  The ABX thing was a quick one, but I wouldn't buy it.  Begging the question?  Much as that argument discusses logical fallacy, what I'm more concerned of is the logic behind the "if you can ABX the diff...otherwise" question.  The motive behind your question is actually clear and is valid, but it is a wrong response to the question at hand, and maybe a right answer to another question.

@odyssey I wonder the question to me "why ever expect a lossy encoder to encode inaudible frequencies?"  Who was expecting?  Certainly it wasn't me.  Is it the LAME devs, because they set the lowpass of 320kbps at 20.5?

Why not add the --lowpass 16000? (Then someone would reply, leave the switches alone)  Now that's something getting everyone in circles.
"Listen to me...
Never take unsolicited advice..."

LAME applying 16kHz cutoff at 320K CBR

Reply #27
The LAME devs tried to make reasonable tradeoffs between available bits and lowpass frequency in such a way that for most people the loss in accuracy due frequency cutoff is approximately equal to the loss in accuracy due to limited bits. Thus the more bits you have available, the higher frequencies you will get.

This doesn't mean that these are the optimum tradeoffs for everyone. Personally, anything above 12 kHz would be wasted on me because I can't hear that high. In fact, I would venture to say that there are many people in my situation, where a lower lowpass than the default would be a better choice.

That said, the next time I rip my collection, it will be to lossless, because I can now easily afford the disc space.


LAME applying 16kHz cutoff at 320K CBR

Reply #29
Shouldn't it?

Not necessarily, no.  In this case the answer is clearly no, it should not.

Is that what was called begging the question?

Your stating that it should as fact is an erroneous assumption that is begging the question.

Perhaps when you design your own mp3 codec you can make sure that it spends bytes on data that people will not hear.  Hopefully it won't come at the expense of compromising parts that people can hear.  Perhaps you'll get nice spectral plots like Blade without the glaring artifacts which often exist even at 320kbits with that codec.  In the meantime I'd not second guess those the Lame developers unless you can cite specific examples where a lack of high frequencies that you feel should exist in the encode are responsible for it not giving transparent results.  Well scratch that, you can second guess them all you like, just don't expect that people will simply take your word as gospel.

The motive behind your question is actually clear and is valid, but it is a wrong response to the question at hand, and maybe a right answer to another question.

...as if you had the right response to the question at hand. 
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

LAME applying 16kHz cutoff at 320K CBR

Reply #30
Here's a 7 second clip from the end of the track in question: http://www.hydrogenaudio.org/forums/index....showtopic=75486 (24-bit FLAC).

Thanks for the clip.  I really wouldn't be surprised by the output you're getting.  I think the explanations cited by JAZ and Dynamic contain the answers you seek.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

LAME applying 16kHz cutoff at 320K CBR

Reply #31
Here's a 7 second clip from the end of the track in question: http://www.hydrogenaudio.org/forums/index....showtopic=75486 (24-bit FLAC).

Thanks for the clip.  I really wouldn't be surprised by the output you're getting.  I think the explanations cited by JAZ and Dynamic contain the answers you seek.


Yep, thanks for the answers.

I just thought it was interesting.

LAME applying 16kHz cutoff at 320K CBR

Reply #32
...Personally, anything above 12 kHz would be wasted on me because I can't hear that high. In fact, I would venture to say that there are many people in my situation, where a lower lowpass than the default would be a better choice...


Is there a way the user can "lower the lowpass" when using LAME?  This could be great if the result is that I can then select more compression and not lose any audible quality that I'd notice.  (I suppose my kids would then think all my music sounds "flat").

If there is not a lowpass setting, then is there a tool to run the audio through prior to LAME?

LAME applying 16kHz cutoff at 320K CBR

Reply #33
...Personally, anything above 12 kHz would be wasted on me because I can't hear that high. In fact, I would venture to say that there are many people in my situation, where a lower lowpass than the default would be a better choice...


Is there a way the user can "lower the lowpass" when using LAME?  This could be great if the result is that I can then select more compression and not lose any audible quality that I'd notice.  (I suppose my kids would then think all my music sounds "flat").

If there is not a lowpass setting, then is there a tool to run the audio through prior to LAME?

--lowpass <frequency in Hz>

LAME applying 16kHz cutoff at 320K CBR

Reply #34
[/quote]
...as if you had the right response to the question at hand. 
[/quote]

Well, me not having the right response does not exonerate you of your wrong "can you ABX" response as well.  I may have a wrong answer, but not rude like yours.
"Listen to me...
Never take unsolicited advice..."

LAME applying 16kHz cutoff at 320K CBR

Reply #35
Well, me not having the right response does not exonerate you of your wrong "can you ABX" response as well.  I may have a wrong answer, but not rude like yours.

Well for once, I don't think greynol provided a rude answer. The answer was already given on the first page of this thread - You didn't understand that, so he had to cut into pieces what the consequences would be for a developer of a lossy codec. The blade example was great. AFAIR it doesn't have a psy-model at all, and thus your spectrograms should look nice - Give it a go.
Can't wait for a HD-AAC encoder :P

LAME applying 16kHz cutoff at 320K CBR

Reply #36
Rio, the "can you ABX" was pointed at you young padawan, not the OP.  You seem to think that you know better than the Lame developers and that the codec is somehow broken.  Provide some evidence to the effect before telling people how things ought to work.  If you feel that I'm being rude, so be it.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

LAME applying 16kHz cutoff at 320K CBR

Reply #37
I have a recording from cassette tape in FLAC. From the spectral I can see it has frequencies up to 21kHz, but when I encode it to MP3 (V0 or 320K) with LAME, there's a clear cut-off at 16kHz (except for at the very end of the track when there's no music, just tape hiss, there's no cut-off here).

I assume this is normal behaviour, I'd like to know why this happens.

I can post a clip / spectral screenshots if needed.


I got this from Mike Giacomelli, if that's of any help here:

http://wiki.hydrogenaudio.org/index.php?ti...cal_information

LAME applying 16kHz cutoff at 320K CBR

Reply #38
@chrizoo, that table is not updated (last version was 3.97, not applicable for 3.98.x), plus the fact that the lowpass filters there for -V1 and -V2 settings are identical (I think it was a typo since they should have different settings).  The discrepancy started at -V5, which should not be 16538 Hz - 17071 Hz.
"Listen to me...
Never take unsolicited advice..."

LAME applying 16kHz cutoff at 320K CBR

Reply #39
@chrizoo, that table is not updated (last version was 3.97, not applicable for 3.98.x), plus the fact that the lowpass filters there for -V1 and -V2 settings are identical (I think it was a typo since they should have different settings).  The discrepancy started at -V5, which should not be 16538 Hz - 17071 Hz.

thanks for pointing that out. Do you have a link to up-to-date material you could post here ? Or the correct values ?

 

LAME applying 16kHz cutoff at 320K CBR

Reply #40
thanks for pointing that out. Do you have a link to up-to-date material you could post here ? Or the correct values ?

I'm sorry I don't have a link for the updated values.  It so happened that while LAME is encoding, I noticed the lowpass filter in the DOS box is different than that in the posted table (for LAME 3.97 on the affected -Vx setting).

For LAME 3.98.x, it has a different setting vis-a-vis to it's 3.97 -Vx counterpart. It is somewhat higher, and may need to adjust the -V setting (ex. -V5 for LAME 3.97 is approximated to -V5.7 in 3.98.x to have the same filter at 16kHz).
"Listen to me...
Never take unsolicited advice..."

 
SimplePortal 1.0.0 RC1 © 2008-2020