MP3 encoding

at0m · Post by **at0m** » Thu Jun 13, 2002 3:59 am

What's your favorite mp3 encoders? Why? I've noticed differences in bass response, Kensuguro complains about losses in stereo field. Which encoder keeps your audio more close to the original wav file?

krizrox · Post by **krizrox** » Thu Jun 13, 2002 4:37 am

Curious about this myself. Where is kensuguro's comment on the subject? Looked but couldn't find anything.

I am using both MusicMatch Jukebox and also tripleDAT for MP3 encoding. But I have to admit I never did a comparison between converted files to see if there are imaging problems.

Which leads me to a somewhat related yet different question. If anyone knows the answer to this, I'd appreciate a response.

A long time ago I had a Soundblaster installed in my PC. I found after I removed it that some apps (like Musicmatch) didn't function properly anymore. What I think I know is that these apps are looking for some soundcard drivers that use the Win Mixer or Volume Control features. Since Pulsar apparently doesn't use these features, Musicmatch (for example) will simply crash if you try to access any functions that utilize those features. Anyone else seen this? Is there a solution other than installing a Soundblaster again?

spacef · Post by **spacef** » Thu Jun 13, 2002 5:29 am

Myself i do a special "Mp3 equalization" of my mixes before converting (time and tests....). No standard settings, i just adapt the equalizing to get the best result, song by song.

kensuguro · Post by **kensuguro** » Thu Jun 13, 2002 5:58 am

krizrox, I commmented about the stereo field in my choir tune.

I'm not sure if it's the mp3 algorythm itself, but I find the stereo space narrowing down after converting. Not a problem if the mp3 is merely a preview, but when mp3 is the format for distribution, that's a problem.

at0m · Post by **at0m** » Thu Jun 13, 2002 6:09 am

Ken,
could some kind of 'anticipation' like SpaceF does w EQ's help you out? Some extra stereo widening before coding or something? Do you use real stereo mp3 or combined?

ernest@303.nu · Post by **ernest@303.nu** » Thu Jun 13, 2002 6:12 am

LAME (Lame Ain't an Mp3 Encoder) is by far the very best MP3-encoder!!!! It easily beats all MP3-competition. Don't take my word for it but check it out yourself, download it here. More information available here.
For non-UNIX-geek-operation I strongly recommend to use LAME with a front-end, there are many at the first link. There's also a .DLL available that allows LAME to be used from within numurous MP3-software packages for the Micro$oft Win-environments.

LAME is a multi-platform open-source project, with loads of people constantly improving the compression algorhythm, and updates appearing very regularly.
Although the constant-bitrate compression outperforms even the latest Fraunhofer-engine, LAME really shines in VBR and ABR encoding. LAME-encoding is fully compatible with most decoding-codecs from about the last 3 years (including those in Wavelab, Soundforge, WinAmp etc)

For more technical backgrounds, discussions about results and information about recommended settings (and there are a LOT with LAME!

) check the forums r3mix.net, especially the technical/comparison forums.

Hope this helps..... after all, the best things in life are free

[edit] I found little point in pre-emphasizing the stereo-spectrum or EQ using LAME on higher (>140kbps) bitrates; the LAME compression is very accurate and does these things as part of it's anticipation of the MP3-artifacts. But for lower bitrates (<128kbps) I like to do some harder compression (OptiMaster!

) so I can have more details encoded, together with a 16-18kHz-lowpass filter within LAME to 'save' the high-freq-bits and use them for better stereo separation and details in the rest of the frequency spectrum. [/edit]

<font size=-1>[ This Message was edited by: ernest@303.nu on 2002-06-13 07:24 ]</font>

subhuman · Post by **subhuman** » Thu Jun 13, 2002 8:12 am

It's all about LAME.

spacef · Post by **spacef** » Thu Jun 13, 2002 8:57 am

I frgot to say i use audioactive because it's simple. I just didn't find how to diasable that awful sound that happens when it finshes encoding.

spacef · Post by **spacef** » Thu Jun 13, 2002 9:04 am

About pan width : i found that a lot of "space" in the mix can be detremined by the higher frequencies. May be it's not automatically this way, but that's what i found.
If you tried the "sfHifi 2 Movie edition" device, there's a knob named "bright" that is actually selective on higher frequencies (from 9000 and more if i remeber well).
By adding this parameters, i have an impression of width, better than in the original file (it's a trick : all the higher spectrum becomes more present, and added to the original sound input, and this gives an impression of widened pan. it's cool on movies especially because a lot of special effects (ambience noise etc) are in the higher spectrum) I haven't tested on mp3 encoding but i'm sure it is very very bad

So i guess that when you encode, you delete higher frequencies, and this is capable of lowering pan space impression.
just a feeling, nothing technical...

ernest@303.nu · Post by **ernest@303.nu** » Thu Jun 13, 2002 9:45 am

Yeah, subhuman, the best things in life are LAME

SpaceF: indeed, human spatial perception is more sensitive for the mid and higher frequencies. Unfortunately emphasizing the hi-freqs doesn't usually trick an MP3-encoder in improved channel separation without a cost

If the MP3-encoder preserves more bits for the (boosted) higher frequencies (as they need to be more apparent during decoding), consequently there will be less bits available for other encoding-tasks, like channel-separation, the rest of the frequency-spectrum, transient accuracy etc, resulting in (slight) artifacts.... Same applies to additional stereo-widening before the actual encoding.
Note that this is only theory, and applies to faithful reproduction compared to the original. Of course you can apply some pre-emphasis tricks to have the MP3 sound fatter or 'better' during reproduction, as there's usually more need for 'good' sounding reproduction in lower-bitrate mp3's than for spectral and dynamic similarity to the "original"

I always do these things for 128kbps mp3's....
As always, let your ears be the prosecutor, judge and jury.....

[edit]typo's[/edit]

<font size=-1>[ This Message was edited by: ernest@303.nu on 2002-06-13 10:48 ]</font>

spacef · Post by **spacef** » Thu Jun 13, 2002 10:03 am

Yes ernst, I Think it's likely to cause phases or agressiveness.

Another thing to try is to encode your mp3 normally, then import the mp3 in soundforge or the like and eq a bit where it can be, then reconvert

I just had this idea but well, i'm not specialist in mp3 encoding (although i made 3 years ago a study for an encoding company, and it appeared that eqing before compression could give a better mp3, but it was for streaming, ie lower quality encoding, that could bear aliasing. Off topic : Eqing in anticipation can also give good result on real audio.