Prologue
The following was done at a time when the state of encoders was in flux. Nowadays, they have matured quite and bit and have consolidated somewhat. I use Lame exclusively, with the --preset standard option, on a Linux box, or 128kbps AAC on a Mac for my iPod.
Abstract
Something that I have an active interest in by virtue of running this Shoutcast is the difference in performance of various mp3 encoders. Most that I have used do quite fine when encoding at 128kbps, a "standard" of sorts. To my decidedly non-audiophile ears, the degradation from "CD-quality" is slight and I have to be paying attention to even notice.
A better test of an encoder, I've found, is the result when encoding at 56kbps. At this level of compression, the various encoders out there really start to differentiate themselves and it starts to become obvious that for real quality, you have to pay. By and large, the "free" encoders produce miserable results at 56kbps but the commercial ones fare better, presumably since they paid the Fraunhofer Institute for their closed code.
Nevertheless, there are some commercial encoders that blow on this test and one sort-of-free one that does really well. Check it out...
Methods & Materials
A 10-sec clip (the first 10 seconds of Talulah Gosh-Break Your Face) was ripped from the CD and saved in wav format. This file was then used as input to all of the encoders listed, except for mp3encoder, which only accepts aiff format. All of the listed encoders were configured to generate 56kbps mp3 files.
Results
| Encoder | Platform | License | mp3 (bytes) | Rating |
| Original wav file | 884,780 | |||
| Bladeenc 0.81 | Linux | Free (LGPL) | 35105 | * |
| Lame 3.13 | Linux | Free [1] | 35271 | * |
| Mpecker 1.0b25 | MacOS | Free [2] | 35555 | * |
| SoundJam 1.0 | MacOS | $39.95 | 37085 | ** |
| SWA 6.0 | MacOS | Free [3] | 36022 | *** |
| Audiocatalyst 2.0 | MacOS | $29.95 | 34377 | * |
| Audiocatalyst 2.0 | Windows | $29.95 | 35236 [4] | *** |
| MP3 Encoder/Media Cleaner Pro | MacOS | $99 | 35854 | *** |
| Bladeenc 0.82 | Windows | Free | 35104 [4] | * |
| CDex 1.10 (bladeenc.dll) | Windows | Free | 35104 [4] | * |
| CDex 1.10 (fraunhofer.dll) | Windows | Free | 34718 [4] | *** |
| Mp3Box 1.2.3 | Windows | Free | 34988 [4] | * |
| Nexgen Encode Studio 3.0 | Windows | Free | 34988 [4] | * |
| Plugger+ 0.4 | Windows | Free | 35176 [4] | * |
| GoldWave 4.11 | Windows | $40 | 35642 [4] | *** |
| Sound Limit 2.51 | Windows | $30 | 35105 [4] | * |
| Audioactive Production Studio Lite 1.5.4 | Windows | $59 | 35490 [4] | *** |
| MusicMatch Jukebox 4.4 | Windows | Free (?) | 34377 [4] | ** |
[1] LAME is actually a patch to the dist10 source distributed by Fraunhofer.
[2] Due to legal entanglements, MPecker was pulled from distribution :p.
[3] SWA (Shockwave Audio) is actually a plug-in for Macromedia's SoundEdit. This plug-in can be used in the absence of SoundEdit via the free wrapper app MP3 Encoder
[4] Courtesy of Ken S.
Ratings: *=terrible **=not bad ***=good
Discussion
No surprise: use the Fraunhofer code, you get quality. The encoders rated "good" were all pretty evenly matched, although I would have to give the edge to the SWA encoder and AAPS as tops. SWA is actually free (see the note above), which makes it better on those grounds, but it's dog slow.
At the other end of the spectrum were the freeware ones (except SWA), which did very poorly. The ones using bladeence were far and away the worst: the results are a garbled mess. The others have some degree of shimmering that I find intolerable.
There's not much middle ground. I guess I'd give a thumbs up to MusicMatch for producing decent results while being free, but thumbs down to SoundJam for producing only decent results at $40.
Another thing to note that this is really a poor "real world" test. I'm only using a 10-second clip of one song here and the results might have been different if I'd chosen a different clip or even a different style of music. For a much more in depth technical discussion and testing, I recommend the MPEG Audio News site (it's Mac-oriented but does have a lot of general technical stuff).
Finally, I encourage you guys to test out your encoders of preference with my sample file and submit the results. Just remember to do your encoding at 56kbps!
A big thanks to Ken for the comprehensive work on the Windows side of things!
