[LAU] What is the best MP3 encoder?

Marius Janßen mariusjanssen at gmx.de
Thu Mar 28 16:14:04 UTC 2013


Am 28.03.2013 16:10, schrieb Jostein Chr. Andersen:
> Hi,
>
> I do have what I consider as a big problem with MP3 encoding, and that 
> is artifacts and degenerated sound quality. End results are things 
> like stronger esses, more "noise" (IE. something like schhhhhh upper 
> overtones on disted guitars) and a loss of clarity. The thing is that 
> a finished 320 kbps MP3 should not sound (significant) different than 
> an original wave file IMO. Note that this artifacts are usually quite 
> subtle, but very audible for the person(s) that recorded it and the 
> mixer.
>
> I usually do mixes in 24/44.1 and I'm using dithering when exporting 
> to a 16-bit stereo file. The artifacts, especially stronger esses are 
> more audible without dithering and it should really not matter.
>
> I'm using Mixbus (Ardour) for exporting the wav files to 16-bit waves 
> and Audacity for converting the file to a MP3. If I upload a lossless 
> file to soundcloud, then the same problem comes there when Soundcloud 
> converts it to it's player.
>
> Unfortunately, I don't have a raw 24/44.1 file and a MP3 I legally can 
> post right now. But I hope that someone have suggestions about how to 
> make the best possible MP3 files from a wave file. :-)
>
> Thanks,
> Jostein
>
>
>
> _______________________________________________
> Linux-audio-user mailing list
> Linux-audio-user at lists.linuxaudio.org
> http://lists.linuxaudio.org/listinfo/linux-audio-user
>
Hello Jostein, hello list,

I have Bobby Owsinski's Mastering Handbook here, and it says lame is a 
very good (or the best!) encoder for generating mp3s.
The codec itself has sometimes some issues with "overcompressed" music 
as well as music with too much high frequency content.

His recommendation is to cut off some very high frequencies to sort that 
out.
You will have to equalise to get the other frequencies encoded better.
Make the signal as hot as you can get it (up to 0dBFS by normalizing, if 
you want it to) but let the codec do its work for last dynamics.
Sometimes more compressed (but not overcompressed) content will get 
encoded better, but let the codec do some work.
You might want to experiment a bit as well.
A slight use of a multiband compressor will do fine but let the codec do 
the rest. Let the codec do its work at "best quality" settings.
Do not waste bandwidth: some songs would sound better at 32kHz because 
the encoder algorithm can focus on mid frequencies.

Coding into mp3 makes the end result hotter than the original. Keep the 
signal level of your source material that will be encoded to mp3 at 
-1dBFS instead of the usual -0.1 or -0.2dBFS so that you don`t get 
digital overs in the mp3.
Regarding bitrate mode he recommends variable bit rate (VBR).

These are just my 2 cents...

God bless, Marius



More information about the Linux-audio-user mailing list