Is there a way to compress without the lowpass filter? I'd like to hear the effects of the compression by itself.

I think you'd use LAME --lowpass 20

I think LAME -k will turn off all low and high pass filtering.

You can't really hear the effects of the compression, because the filtering is part of what makes the compression work properly. By turning off the filtering you are by definition stopping the encoder from working properly.

To use the -k switch, which turns off lowpass filtering, you would need to use an older version of lame. Latest version doesn't support it.

Also, even if the highest frequencies are not filtered out doesn't mean they will be encoded.

Oh, I didn't know that! So there's no way to disable the filters? When I use MP3, I use the presets, which I believe have finely tuned settings for everything including the filters, but just for experimental/testing purposes filters can't be turned off?

I'm not sure I understand what you are trying to achieve, but if it is to demonstrate how badly mp3 artefacts could possibly be, you might try to find some old encoders - i.e. Xing encoder and try lower bitrates.

If I misunderstood your intention, please elaborate what you are trying to do.

Pretty much just that.

And I'll give it a try, thanks.

I would like to know 2 things:

1. Is it possible to use a lowpass of 50 kHz? Of course, this is completely useless or downright damaging to the quality of the MP3, but I'm just curious.

2. Let's say someone establishes he/she cannot hear frequencies highet than 17.5 kHz. If he/she chooses --lowpass 17.5 as an additional option which overwrites the lowpass LAME would choose without it, how are then those extra available bits used? Will that in any way improve quality?

1. What encoder, what version? If you ask about LAME 3.98, then no: max. lowpass value is 20.5 kHz.
2. Yes, it should improve quality (CBR/ABR) or reduce bitrate (VBR).

There's no way to get more than half the sampling frequency.
So with music originating from CD (44.1 kHz sampling frequency) the theoretical limit is 22 kHz, and in practice it's a little bit lower due to the width of the necessary lowpass filter of any digital audio system.

Everybody with the 20 to 20000 Hz theoretically optimum frequency range in mind is best off sticking to the Lame defaults. These respect the limitations of the quality setting chosen and allow for best highest frequency behavior within this restriction.

Everybody who's made the experience that he doesn't benefit from extremely high frequencies can use a lowpass according to his needs due to the flexibility of Lame. He will benefit from the advantages lvql described. Of course it's a personal question whether or not it's worth while.

It is also important to know that hearing single tones up to a certain frequency is very different from hearing actual music. I know many people who say that they can hear up to 20KHz when it comes to "music." However, they are just testing singular tones. I then conduct further tests and their ears cannot go beyond around 16KHz when it comes to songs. Songs mix lower range material in with higher range information. Ears often have a hard time picking out a high range when they are fed something across the spectrum like that. So make sure that you use real music (unless your music consists of simple tones) when trying to determine what frequency range you can actually hear.

Are there even any often-used musical instruments that can reliably go above 16 KHz? I recall reading about a special custom-made pipe organ that could reach as high as 17 KHz, but I doubt it made an appearance in more than a few songs.

Many forms of electronic music rely on frequencies in that range for a sense of "crispness," or else they simply use a heavy boost around 16 kHz to create a psychoacoustical effect which fakes the appearance of frequencies above that range.

They're quite easy to detect on a decent set of headphones, just listen to the very top end of the song... the fuzz of the snare, the sizzle of the hats... it all lies around 16 kHz or even slightly above.

http://www.its.caltech.edu/~boyk/spectra/spectra.htm


C.

Interesting findings about real world musical content above 20 kHz.

Anyway we do need a lowpass for CD music which is roughly 21 kHz at its highest possible implementation.

In the light of the article the question is: 'if we can't have the energy above 21 kHz which can be pretty strong on occasion, does it matter if we go a little bit lower and don't use anything beyond say 17 kHz?'

Which brings us back to where we were. Everbody thinking in terms of 'the audible audio range is 20 to 20000 Hz' is best off with the default Lame settings. And those who are satisfied with a lower lowpass do it according to their needs.

it all comes back to "can it pass an ABX test"

Hello,
I've been using the mp3 converter from AIMP2 for a while, because I think it's quite convenient, but today I realised that when chosing the best quality lame encoding (vbr dual mono), it was making a lowpass at about 12KHz !!!
Personnally, I don't hear much above 15KHz, but 12KHz is a bit short, isn't it ?

These are the settings in encoders.ini:

[MP3]
App=lame.exe
Ext=.mp3
ForceTags=1
Presets=ABR [Joint Stereo];ABR [Dual-Mono];CBR [Joint Stereo];CBR [Dual-Mono];VBR [Joint Stereo];VBR [Dual-Mono];
ABR [Joint Stereo]=--abr %br% -m j - %out%
CBR [Joint Stereo]=-b %br% -m j - %out%
VBR [Joint Stereo]=-b %brmin% -B %brmax% -V %q% -m j - %out%
ABR [Dual-Mono]=--abr %br% -m d - %out%
CBR [Dual-Mono]=-b %br% -m d - %out%
VBR [Dual-Mono]=-b %brmin% -B %brmax% -V %q% -m d - %out%
Q_PARAMS=0;8;4;High;Small;

I replaced lame.exe with 3.98 and changed the bold line to

VBR [Dual-Mono]=-b %brmin% -B %brmax% -V 0 -q 0 -m d - %out% --lowpass 20

and the lowpass disappeared, but I was wondering if these settings were consistent with the fact that high frequencies are now kept.

What do you think would be the best settings without lowpass ?

I usually use it to compress Wav 48Hz 24bit recordings, and for sure I would like to have the best settings possible.

These are by far the best settings:
Lame.exe -a --abr 320 -b 32 -B 320 --bitwidth 32 --clipdetect --cbr 320 --comp 1 -e c -f -F --freeformat -h --highpass 0 --highpass-width 0 --lowpass 24 --lowpass-width 0 -m d --noreplaygain --nores -p --preset insane --priority 4 -q 0 --replaygain-accurate --resample 48000 --scale 1 --strictly-enforce-ISO -V0 --vbr-old -X7

Delete everything except -V 0 and %out% and you should be OK. Everything else in there will potentially lower the quality.

I hope this is a joke?

Sorry, i could not resist. I justed wanted to say that more switches is not better. If you don't know what a switch does then don't use. I totally agree with using -V only.

As previously pointed out by pdq, you need to change some things. You don't want to use dual-mono encoding, you will want to change that to joint stereo. Additionally, the use of --lowpass 20 is not needed as Lame pretty much uses that lowpass at -V 0 anyway. Take out -q 0 as it is not needed either. I believe that Lame 3.98.2 is ignoring that command as -q 3 through -q 0 will use the same "quality" setting. Mark7 made the point that adding more switches is not better, I full heartily agree. The Lame developers have finely tuned each VBR quality setting for optimal performance. Often times, the minimal command lines (such as using just -V 3, -V 0, -V 2, etc.) produce the best results. There are some people who take the time to determine the proper lowpass settings that they need through proper testing. That is perfectly fine. However, most of the time, people use long command lines when they don't really know what they are doing.

They will read a guide online suggesting to use some mile long command line as 10000 self-proclaimed audiophiles will praise it. Follow the Lame guides in the Hydrogenaudio wiki, they were put there for a reason.

Anyone who feel that some techniques used by lossy encoding, i.e. joint-stereo in any way is bad for the listening pleasure, should stop bitching and use a lossless format instead.

Correct me if I am wrong but don't some lossless formats, like FLAC, use a technique similar to joint stereo for encoding?

You shouldn't say so quickly that it was a joke, I was going to copy-paste

I used the short settings with joint-stereo, now the lowpass is back to 19.5 KHz and the file is not smaller than with my settings (which helps me a lot for accepting that short command line).

I don't use lossless because the file is 5x bigger, which would fill my HD too quickly, but with the increasing size of storage and bandwidth, one can wonder if compression has a future ...

Thanks for your answers !

Yes, but a form of lossless joint stereo that was also very technically explaned in a recently thread. mp3 joint stereo is not really lossless.


This is just not a valid point anymore. I have ~35.000 tracks in my library and I have not even filled my 1.5TB drive yet, which cost roughly the same as 6 CD's!