Skip to main content

Topic: the use of "-k" switch in LAME 3.97a11 (Read 5848 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • Blenvid
  • [*]
the use of "-k" switch in LAME 3.97a11
Hey everyone,

I've been exhaustively reading all of the posts here and on the net this week trying to find an answer for this, but I haven't found one. I'm new to the ripping of cds with EAC (version 0.95) and encoding with LAME (version 3.97a11), but have been ripping music to create mp3s since 1997.

I noticed while encoding some mp3s with EAC and LAME that LAME said it was using a lowpass filter. I was simply using the preset "--preset standard" in EAC to pass to LAME. I then noticed that all of the mp3s I ripped with this preset used the lowpass filter. It's called the "polyphase lowpass filter."

Just to include the defenition of a lowpass filter for those that don't know what it is, it's a filter that attenuates (reduces in level) the frequencies above its cutoff frequency.

I looked around the web and found a couple examples of mp3s that were ripped with a LAME encoder at lower bitrate settings and there were some very high frequency artifacts which were created by the encoding process. I'd imagine the lowpass filter was included to reduce one's ability to hear these artifacts.

If you take a look in the LAME "Full command line switch reference" document included with LAME, it says under the "-k" switch that the Full Bandwidth setting "tells the encoder to use full bandwidth and to disable all filters. By default, the encoder uses some highpass filtering at low bitrates, in order to keep a good quality by giving more bits to more important frequencies. Increasing the bandwidth from the default setting might produce ringing artefacts at low bitrates. Use with care!"

Why would anyone want to put filters on their music if they're using the LAME "--preset standard" setting which isn't even a lower bitrate setting, but has bitrates in the 170 to 210 kbps range? 

To test, I went ahead and put "-k" in my command line; the resulting command line looked like this, "--preset standard -k" without the quotes of course. I want all of my mp3s to be in Full Bandwidth, not with filters. I noticed that the files with the -k switch sounded a little more open, more transparent, a little more like the original, yeah!!!

What's the word from anyone else on LAME's default filters using the "--preset standard" preset? The LAME recommended settings everywhere don't even mention this. Because we're using a higher bitrate, wouldn't you want your mp3s to be at Full Bandwidth as well, without any filters? I've noticed that some people like to put bass boost filters on, but people who just want the music as close to it was on disc? Why aren't people using the -k switch? 

Thanks for any input.

-Blenvid

  • kwanbis
  • [*][*][*][*][*]
  • Developer (Donating)
the use of "-k" switch in LAME 3.97a11
Reply #1
Quote
I noticed that the files with the -k switch sounded a little more open, more transparent, a little more like the original, yeah!!!

you should take a look at TOS#8

  • stephanV
  • [*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #2
Whats "more" transparant? It either is or it isn't.

I think if you don't include listening test to support your findings, not many people will pay much attention to this. If -k would obviously be a quality improvement, it should have been included in the preset don't you think?
  • Last Edit: 17 August, 2005, 02:08:04 PM by stephanV
"We cannot win against obsession. They care, we don't. They win."

  • Blenvid
  • [*]
the use of "-k" switch in LAME 3.97a11
Reply #3
Thanks for the quick replies.

I'll take a look at TOS#8 Kwanbis, thanks.

Quote
Whats "more" transparant? It either is or it isn't.

I think if you don't include listening test to support your findings, not many people will pay much attention to this. If -k would obviously be a quality improvement, it should have been included in the preset don't you think?
[a href="index.php?act=findpost&pid=320741"][{POST_SNAPBACK}][/a]


Oh, it wasn't my intention to imply that the use of the -k switch was better or worse for everyone else stephanV. I was just wondering what other people thought or knew about this, if the filter was necessary for example, or if they too had the same findings, or if they read it didn't make any difference either way, etc....

Just to comment on your comment about transparency though stephanV, varying degrees of transparency can be achieved by the inclusion or exclusion of frequencies which were obtained in an original recording. For example, if a choir was recorded in a church, but the frequencies which emphasize the acoustic space of the church (reverb) have been filtered out, the recording would sound less transparent, more flat, less alive, as compared to the original.

I just want to get a close as a copy to the originals as possible, without going lossless of course.

-Blenvid

  • evereux
  • [*][*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #4
Quote
To test, I went ahead and put "-k" in my command line; the resulting command line looked like this, "--preset standard -k" without the quotes of course. I want all of my mp3s to be in Full Bandwidth, not with filters. I noticed that the files with the -k switch sounded a little more open, more transparent, a little more like the original, yeah!!!
-Blenvid
[{POST_SNAPBACK}][/a]


It's unlikely that you can hear what you think you can hear. To eliminate placebo it is important to carry out a [a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=16295]blind listening test[/url].

I really wouldn't worry about --preset standards lowpass filtering. MP3 is a lossy process, you will gain losses across the board. In order to save bits that can be better used elsewhere the audio is lowpassed to remove frequencies we're unlikely to hear anyway (at preset standard, at lower bitrates the lowpass is more aggresive).
daefeatures.co.uk

  • sony666
  • [*][*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #5
get yourself a decent wave editor (like an old Cool Edit Pro Demo version), open up your original .wav file and cut out everything under 19000 Hz.

listen to what is left. that usually answers the -k question

Edit: oh, make sure the software does not automatically amplify after applying the highpass (to compensate the loss of material).
Otherwise your tweeters/headphones may get fried.
  • Last Edit: 17 August, 2005, 02:52:59 PM by sony666

  • stephanV
  • [*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #6
Quote
Oh, it wasn't my intention to imply that the use of the -k switch was better or worse for everyone else stephanV. I was just wondering what other people thought or knew about this, if the filter was necessary for example, or if they too had the same findings, or if they read it didn't make any difference either way, etc....

You did however imply that it was better for you, but to have a meaningful discussion it would be nice if everybody using -k can back up with a listening test he is hearing an actual difference. It prevents everybody suggesting their favorite non-sensical commandlines of 10 paragraphs long. Now I'm not suggesting you do not here the difference, cause I really don't know and I don't think it would be impossible to hear, but its more fruitful to discuss something that you can prove to have observed.

Quote
Just to comment on your comment about transparency though stephanV, varying degrees of transparency can be achieved by the inclusion or exclusion of frequencies which were obtained in an original recording. For example, if a choir was recorded in a church, but the frequencies which emphasize the acoustic space of the church (reverb) have been filtered out, the recording would sound less transparent, more flat, less alive, as compared to the original.

AFAIK the term "transparancy" is most commonly used here to indicate that someone cannot hear the difference between an original file and a lossy encoded file from that original (A file is called transparant, when no audible differences with the original can be heard). This of course means that transparancy is quite a personalised quality of a file: not everybody has the same hearing. Keeping that in mind, there is no such thing as more transparant: you cannot not hear the difference in more degree than one. You either do or don't. Now one could argue that some files would sound closer to the original than others, but I wouldn't call that "more" transparant... maybe this is just nit-picking though...
  • Last Edit: 17 August, 2005, 03:24:26 PM by stephanV
"We cannot win against obsession. They care, we don't. They win."

  • [JAZ]
  • [*][*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #7
Quote
If you take a look in the LAME "Full command line switch reference" document included with LAME, it says under the "-k" switch that the Full Bandwidth setting "tells the encoder to use full bandwidth and to disable all filters. By default, the encoder uses some highpass filtering at low bitrates, in order to keep a good quality by giving more bits to more important frequencies. Increasing the bandwidth from the default setting might produce ringing artefacts at low bitrates. Use with care!"

[a href="index.php?act=findpost&pid=320735"][{POST_SNAPBACK}][/a]



Hummmm...  This is a typing error in the documentation. It is not a highpass, it is a lowpass.



Going back to the question, (even when it is expected an ABX for this claims), the use of lowpass filters in LAME is because of a defect of the MP3 specification, which causes high frequencies ( search for SB21 , >16Khz ), to increase notably the use of bits (all bands need more bits to store the same information).
Reducing the highs, allows a better file overall. (better in the sense of more transparent, as bits used more wisely).

It is accepted that, in music, past 16Khz is not hearable in the general case, even when single tones of 18Khz, 19Khz, etc could be heard.
Probably, the content you are encoding is noisy by nature (heavy metal or alike?). Here it could be noticeable, but with carefull listening.


To conclude, if this filtering really matters you ( and it means, you don't imagine you hear the difference), else try --preset extreme (VBR) or --preset insane (CBR, max of what MP3 can do), or forget MP3 althogether.

  • Lyx
  • [*][*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #8
If you want higher quality, then you want to use a lowpass.

Why? Because available bits in mp3 are not infinite - the additional bits which you spent on (barely noticable) ultra-highfrequencies are taken away from much more noticable frequencies - thus, you degrade overall quality by using -k.

If you dont want anything to be "taken away" from the audio, then you should switch to a lossless format - because lossy means that there is always a theoretical loss in quality - the difference is just "where to take something away", not "if".

Dont try to "tune" an encoder. Default settings are the default for a reason.
  • Last Edit: 17 August, 2005, 05:24:37 PM by Lyx
I am arrogant and I can afford it because I deliver.

  • tortis
  • [*]
the use of "-k" switch in LAME 3.97a11
Reply #9
I can't believe people are still having this sort of discussion in 2005!    -k is definitely not going to do ANYTHING you would notice.  Not even if you had hearing like a dog!
  • Last Edit: 18 August, 2005, 02:23:11 PM by tortis

  • Madrigal
  • [*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #10
If anything, it seems reasonable that use of the -k switch should lower overall quality, by using up (wasting) bits on inaudible frequencies, that could otherwise have been used to improve the encoding of audible frequencies.

Regards,
Madrigal

  • MugFunky
  • [*][*][*][*]
the use of "-k" switch in LAME 3.97a11
Reply #11
removing the lowpass would only improve coding on test signals, and then only if you can hear a 19+ khz tone in isolation.  the highest i can hear is 19700 Hz, and that was measured with a good soundcard on very very good headphones (i could only get up to 19500 on my previous pair).

and that was with the volume just below the threshold of distortion - ie maxed to it's usable limit.

but put in some natural sounds (or even low level pink noise) and that tone would be completely masked.