Skip to main content

In preparing a track for artist review I ran into a piece that really fell apart in the MP3 conversion from Adobe Audition. I thought it might be a good test track for different MP3 encoders.

I exported the same WAV file out of AA, Samplitude, and then used LAME. I selected the highest quality option in all cases, 128 kbps CBR. I have links here to the original WAV file and the three resulting MP3 files. IMHO, Lame really shines on this one.

FYI

[="http://www.cheap-tracks.com/mp3/MP3_Test/MP3_Torture_Test.wav"]MP3_Torture_Test.wav[/]="http://www.cheap-tr…"]MP3_Torture_Test.wav[/]

[[url=http://="http://www.cheap-tr…"]MP3_Torture_Test_AA_CBR_128.mp3[/]="http://www.cheap-tr…"]MP3_Torture_Test_AA_CBR_128.mp3[/]

[="http://www.cheap-tracks.com/mp3/MP3_Test/MP3_Torture_Test_Sam_CBR_128.mp3"]MP3_Torture_Test_Sam_CBR_128.mp3[/]="http://www.cheap-tr…"]MP3_Torture_Test_Sam_CBR_128.mp3[/]

[[url=http://="http://www.cheap-tr…"]MP3_Torture_Test_Lame_CBR_128.mp3[/]="http://www.cheap-tr…"]MP3_Torture_Test_Lame_CBR_128.mp3[/]

Tags

Comments

anonymous Mon, 09/11/2006 - 18:30

As a guideline, if you plan on real-time streaming then use CBR, else use VBR. VBR allows the encoder to dedicate more bits during "troublesome" parts, but can overflow input buffers when streaming.

I didn't listen to the mp3's. but I gave a quick listen to the wav file. One of the original quality tests for the MPEG audio encoders was a recording of castanets. Precussives contain a surprising amount of high frequency content, which lossy encoders don't handle well (eg, swishy cymbal sound).

anonymous Mon, 09/11/2006 - 20:23

Sounds pretty interesting.

I am not, by any means, an expert at this but.... of all the mp3 encoders I've ever tried (Including Lame) I got the best sound ouf of the Sonic Foundry mp3 encoder included in sound forge.

I always encode at 192kbps CBR (some people say it's overkill). The thing I dislike about most mp3 encoders is that garbled sound in the higher frequencies (specially on cymbals). I never had this problem with the aforementioned encoder. I'll use it to encode the wavefile you posted using the same profile and compare it to the rest.

anonymous Tue, 09/12/2006 - 12:08

Soundforge uses Fraunhofer's encoder - mostly everything is either a choice of Fraunhofer or Lame these days (Blade, Xing etc have fallen by the wayside).

Encoding at 192 is not overkill, it gives a much more faithful reproduction of a .wav file, particularly one with a heavy bass emphasis and much more clarity in th 12-15 Khz range (for cymbals esp.) . 128 was a standard back in the days of the 6GB HDD and slower computers and remains so due to restrictions from some websites (i.e. soundclick's free accounts) that will not allow any other bitrate.

I tested a series of my own mp3's using Voxengo's free SPAN vst spectral analyzer and found in 4 out of 7 recordings Fraunhofer gave a slightly better response and 3 out of 7 Lame - however, if pumping the volume up on mastering, I found Lame a better encoder, Fraunhofer caused some slight clipping. The results were inaudible to me anyway, so I wouldn't recommend one over the other, both are good.

I also find Ogg Vorbis to usually encode slightly better than mp3 but vorbis is a vbr and not supported by many websites, although is the primary choice if uploading to the Internet Archive.