using lame,
how can i enable the encoding of 16Khz+ content ?
please don't ask me why or tell me it's wrong/unnescessary
blah blah blah.. just tell me how.

second question: why can't this "frequency cut-off" during encoding,
be "soley" controlled using a "--lowpass[value]" ?
would'nt that be more easy ?
maybe because all a lowpass does is a "lowpass",
and then the psych-model has a definite saying in what will finaly
stay or be left out ?

i'm realy getting tired-sick of what the psychmodel does most of the time,
doing what it thinks is best,
cutting almost everything at 16Khz leaving most 16K+ freq out
no matter what i do it cannot be steared the way i want to.
it doesnt seem to bother raising the 16K+ freq. content in a mp3
even when a higher bitrate is used. why ?
maybe there's a way but i dont know how.


morelli

What command line are you using AFAIK alt-preset standard leaves up to 18Khz

óh boy.. there's that -aps again.

i feel every track has its own needs of settings in order for it
to sound the way i like it,
or as close as i feel is okay compared to the original.
without the file to be too bloated or too small cutting detail.
so in general i'm using alot of different settings and changing them
in small steps as much i as feel nescessary.
there's not 1 setting i use in particular.

i'm testing to see what -aps does at this very moment.
but i thought (if i indeed tested it earlier) even -aps somehow
cuts and seems to lower (?) the 16+ freq content
somehow.
it even cuts away completely above 16+ if it 'sees fit'.
but i do not know if that was the exact conclusions i had.
i do not use -aps cause i didnt like it when trying some encoding with -aps
in the past

But my question was plein and simple,
is there or is there not a way to enable encoding of 16KHz+ content
, and somehow stear (as far possible) 'how much' highfreq information has to be encoded ?

is that too difficult without bringing up -aps ?

-aps might work for most, but you could also say (logocaly) it wont work for
everybody or in every case.
i realy do like the effort thats being put in such a setting that will
cover the encoding of alot of audio data.
but you'd be stepping toes if you would be saying one setting is
better than all or most other settings.
cause thats the feeling i get regarding this -aps, reading posts from last year.

i'm just trying to get lame to encode my audio the way i like it to.
or at leats try to stear it in a certain direction.
i'm not one of those who desperatly wants someone else to try MY
settings.
i feel perception is something personal.

imho.
lame can be darn difficult to understand with all that tweaking
going around in all those different kinds of thresholds and switches changing from one control purpose to another,
and thresholds being changed all the time.
the ones making and changing those thresholds and switches
and their purpose have their own unique perception as well, i feel.
i hate it when there is so much cutting around and then telling
others what does or doesnt matter. i think it's personal.
perception isnt just numbers, even with extensive abx testing,
i think you cannot soley rely on the numbers and stats deriving
from such tests. even if they are more based upon perception then
numbers and robot measuring.
please let people make up their own minds.

morelli

Last time I did an encode using alt pre std (with LAME 3.92), the command line status showed the rolloff beginning around 18 kHz (alt pre ext may raise this a bit)...well above your 16 kHz. Unless it's a brick wall or chop filter it should attenuate the higher frequencies gradually. If you want to roll your own then look at the html's that come with the LAME zip. You won't get much help here with encoder settings/switches because most people know that your not going to beat the presets for sound quality, and they (and I) have better things to do.

xen-uno

well, since you are so stubborn:

lame -k --Y0()r-l33t-(0/\/\./\/\4|\||)-l1n3-G035-h5r5

will keep ALL frequencies.

i'm not pro -aps so now i'm stubborn ?
i have a question and -aps is THE only answer ?
-aps isn't everything.
a while ago i tried -aps and din't like it is that a crime ?

-k isnt the answer i'm lookin for either.. tried that before.

lame html's wont tell me much more about the workings of some switches.

then what do you want.

-k will keep ALL the frequencies. your request was: 'how can i enable the encoding of 16Khz+ content ?' your answer is -k.

[HUMOUR]
No, it's not the ONLY answer, there's also 'alt-preset extreme' and 'alt-preset insane'.
[/HUMOUR]

Well, you seem to be "demanding" an answer from folks that are here on a "volunteer" basis......I don't get it...

but anyhow...

What version of LAME were you using when you reached the conclusion that "a while ago i tried -aps and didn't like it....". There have been many versions. 3.90.2 (considered the best), 3.92, and 3.93.1 (has come around as well) are all considered solid in regards to the presets. Try an encode again with any of these.

xen-uno

Musepack is the answer

There are so many people on HA these days who insist on answering their own questions. Post the question->Ignore the answers Don't make no sense to me.

If you are not happy with the your mp3 files or --aps it might be because mp3 is not capable of delivering the quality you demand. Try lossless or write your own mp3 encoder

Just to further muddy the issue, if high frequency is really important, and you hate psychoacoustic models messing with your music, try Wavpack lossy, if you don't need mp3 compatibility.

The higher bit rates might not fit your requirements, but Wavpack lossy at 320 kbit has been very impressive with my tests, both as is and as a source for later transcoding for portable use. It's a neat compromise between the lower bitrates of lossy codecs, but without the artifacts and with a more manageable file size compared to lossless. I personally prefer it over LAME aps, ape and api by a considerable margin, and considering api is also 320 kbit, it knocks LAME out of contention for my particular needs, but yours may be different. Encoding is also freaking fast compared to LAME aps.

Give it a spin, you might be surprised. You come across as someone who takes their music seriously, and may be more sensitive to your music's high end. I don't have golden ears but I much prefer the treble in my music through Wavpack. Some of this is possibly from high freq cut off, but I think most of it is the lack of any pre-echo in Wavpack. No more smeared hi-hats!

I have posted some comments and tests regarding my findings in the other audio codecs section of the forum, mainly focusing on transcodes from wavpack/musepack/vorbis/mp3 for Minidisc, but there are some general listening comments there too.

Xenno's comments are valid regarding LAME version, but I'm guessing that if you dislike aps in 3.90, you will also not like it in later versions also, and vice versa.

Den.

--lowpass is controlling the lowpass.
-k is disabling every bandpass filter.

-aps is not everything, but it's simply the best VBR profile available. If it is the lowpassing that annoys you, try using --lowpass option, but I doubt that the default -aps lowpass of 19.5 KHz is audible.

Then you have -api, that is no longer VBR, but 320 Kbps CBR, and some times yields better results. You can combine it also with the --lowpass option (don't remember the default lowpass for -api, maybe 20 KHz?).

If you are not satisfied, then you should try MPC starting from standard profile. If not, satisfied, maybe the WavPack lossy suggested is a good idea. If not, switch to lossless.

...but just one suggestion: try some ABX'ing.

(Edit: corrected lossy for lossless, I just realized at same time than Gabriel)

Means "If not, switch to lossless"

You're never going to get perfect encoding of >16khz content with MP3 due to limitations of the format spec.



Because of the above mentioned limitation in the format spec.



Again, because of the problem mentioned. Do a search for sfb21 and you should find a bunch of information about this.

Speaking of high frequencies:

Some one, some where else, on some other forum stated that even when higher frequencies are inaudible, their harmonics have an effect on other (lower) frequencies. For example, 20 kHz frequencies have a fifth harmonic at 4kHz and that harmonic can affect those note's "timbre".

Any thoughts?

The harmonics go from lower frequencies to higher frequencies generally and still, filtering the high content does not affect the low content. The thing you're talking about sounds more like aliasing that any other thing, and that's NOT a good thing.




@ morelli :

The others have already replied to you correctly, but there's one thing that gave me attention... if you encoded something with --aps , and gave you a "cutoff" of less than 16khz... I quite believe that what you are looking for is an exact reproduction of the spectral view, not caring at all if that's hearable or not.

In that case. I would INSIST that you have just two doors to choose from:

a) educate yourself (ABX, etc).
B) go lossless and stop giving nuts because people suggest you the best that MP3 can do.

If the harmonic was produced by the instrument it would have captured by the recording, and is therefore a noise on the source CD at 4KHz (or whatever) that can be encoded (or discarded if it is not 'significant'). If it's a harmonic from any other source it's interference, and you don't want it to be encoded.

This is apart from the point JAZ makes that harmonics generally go up, not down (something I don't know about)

Cheers, Paul

Been here already ?

http://www.hydrogenaudio.org/forums/index....7516#entry74075

EDIT : recent development : http://www.hydrogenaudio.org/forums/index....t=50#entry96338

Btw, you should know that your hearing ability is quite limited at 16kHz, and even more limited at 18kHz:
http://ccrma-www.stanford.edu/~jos/bosse/A...ld_Hearing.html

That is why after encoding there is less high freqs: many can not be heared

yup...


BWT:
(perhaps daft)

if the suppressing of dynamics.. (grand scale)
is part of the compression algorithm..

wouldn`t it then be a pointer to
further development.. in that area.

i would like my compressed files.. to be as identical
to the original.. as possible..
if it means bigger files..
let it be so..

if the freq. is cut.. at 16000.
the music have losts it`s feeling..

Hopefully, when using --preset standard, the cutoff is higher than 16kHz.

Quick way to check it:
encode a track with "--preset standard" and with "--preset standard --lowpass 16". You can check that the first encode is bigger. If there was a cutoff at 16kHz, they would be similar in size.

bottom line, i'd say, is that he's not gonna get what he wants from MP3, without experimenting for 4-20 hours with his own settings. if he wants decent bitrates at the frequencies he's demanding, i'd say OGG at Q6 to 7 (192-224kbps, lower if there's little hard rock-type distortion and the like kind-of-noise) otherwise, i'd say if he's gonna want to play around with LAME, expect 256kbps or higher type bitrates. you just can't have it all =P

I'd say that, bottom is, any efficient lossy codec is going to remove parts of the spectrum, maybe not always, but when it considers it is useful. That is the base of lossy coding, to remove what you can't hear. It is either that, or switch to lossless.

The only lossy algorithm that I know of that doesn't do that, is WavPack lossy, but it is quite inneficient (bitrates around 320 kbps are needed), and others such as Musepack are regarded as being capable of achieving higher levels of transparency, even when they need lower bitrates.

This is a good point; however if a full spectrum source is played then you could get those harmonics nonetheless and (if the harmonics are present) could be part of the final sound the engineer produced at mastering and, therefore, those harmonics would not be present during playback of lossy source.

No, I'm not trying to advocate the use of higher lowpass filtering but the statement I saw about harmonics did make me curious. Now let me read the link provided before I continue.

It is a really silly thing to be giving so much thought to... keeping high frequencies which you can see but not hear.

But maybe the solution would be to use one of Lame's other ATH's? Sure, something else will suffer, but the users gets some more high-frequencies back.

I've noticed in this thread that a couple people have said:
". . . harmonics generally go up . . ."

Don't harmonics, by definition, always go up?

Thanks, Dex

When a string is set into motion the frequency or "pitch" we usually hear is referred to as the "fundamental" pitch. That pitch is also called the "first harmonic."

The two points which attach the strings are the "nodes" of the fundamental vibration. What is not obvious is that each string also vibrates simultaneously at many other frequencies . . .

. . . each progressively higher in pitch and lower in amplitude (volume).

When the fundamental pitch is created by the vibration of a string of length x, additional frequencies, called overtones or higher harmonics begin vibrating with effective lengths of x/2, x/3, x/4, x/5, etc.

Each of these additional frequencies vibrate between additional imaginary nodes that are located exactly at the points on the string which are integer divisibles of the string length.

The basic physical law which governs vibrating objects states that the frequency of a vibrating string is inversely proportional to the string length. What that means is if a string of length x vibrates at frequency f, than that same string if cut to length x/2 will vibrate at 2f.

Since strings vibrate simultaneously at all integer subdivisions of the string length (effectively creating shorter and shorter string lengths), the vibrating string will produce additional pitches, called higher harmonics or overtones, at frequencies of 2f, 3f, 4f, 5f, and higher.