My Fraunhofer Vs LAME Test Results

[chadbang]

I've been using two encoders trying to make "archival" quality MP3s to load on my iPod. So I decided to test and do a comparison of the Fraunhofer encoder in Quicktime 5 and the LAME encoder, using EAC as a front. The Fraunhofer MP3s were encoded on a Mac and the LAME on a PC. The playback of the MP3s was through my iPod into my home system: Harmon Kardon preamp, Adcom 535 amp, Spica TC-50 speakers.

My first experience with LAME wasn't pleasant. I was not using an alt/presets and I just came up with my own setting. (320kbs/non-VBR). I found the treble too spikey on these first LAME mp3s. When comparing LAME to a Fraunhofer encoded MP3, the Fraunhofer encoding sounded better - smoother and pleasantly rolled off.

Fortunately, Alereon got me set up with alt/preset/extreme. And he also pointed me to an older build of LAME -- LAME 3.90.3. I had been using LAME 3.93, but a quick comparison and listen told me I liked the slightly fuller sound of 3.90.3.

Now I could compare the best of LAME to the Fraunhofer codec. First of all, the spikey treble of LAME was gone and I would say the overall tonal balance of the two codecs was now more even. Both pleasant to listen to, and I even noticed that the vocals were now even smoother with LAME 3.90.3 than with Fraunhofer. There was a slight dryness to the Fraunhofer encoded vocals that hadn't been apparent before. First point the LAME.

It was again a tough call at first - both good codecs - but then I noticed that there was something slightly more musical about the LAME 3.90.3 MP3s. Okay, we all hate that term. It's so hard to define. Musical? So I thought I better think about it and figure out why my toes wanted to tap to the LAME tracks and not the Fraunhofer MP3s.

I was listening the Nick Cave and the Bad Seeds excellent audiophile treat "The Boatman's Call". Recorded at Abbey Road, it is a wonderfully rich, natural recording with lots of ambience. Almost virtually a live performance in its recording technique. It was with this album that I realised WHY the LAME MP3s sounded more musical. The LAME tracks vs the Fraunhofer track had better dynamics.

When various instruments would kick in, there was decidedly a stronger, better defined attack to their appearance. Hence, the music was more involving and more exciting. Hence the overall sound was more musical

So there it is. Sort of a pain and a lot of work, but that's my accessment of the two codecs. For critical listening I prefer the LAME 3.90.3 MP3s for the reasons I've given. Fraunhofer is also very good but I, like so many other Head-fiers, give the edge to the LAME codec.

Cheers.

 

[gpalmer]

Just out of curiousity, which version of the Fraunhofer codec were you using? I used one previously that would only do up to 56 K or 48 K or some other low bit rate, but was free. It really wasn't much of a contest between the two, LAME was a hands down 5 second comparison winner, even when the encoding was adone at the same level. Is Fraunhofer giving away a better one free now? I should have mentioned, LAME had better frequency extension, Fraunhofer cut off the top and bottom. This was around 4 years ago though, so I'm sure a lot has changed.

 

[fewtch]

Quote:

Originally posted by chadbang When various instruments would kick in, there was decidedly a stronger, better defined attack to their appearance. Hence, the music was more involving and more exciting. Hence the overall sound was more musical

What you're probably noticing is less pre-echo with Lame (a particularly nasty MP3 artifact that smears transients -- in particular, instrument attacks).

I was feeling a bit unhappy with how "digital" my soundcard has been sounding, so I went and found a hard-to-find plugin for WinAMP (SSRC resampling) and am listening to all my music files (compressed or uncompressed) upsampled to 88.2/24 with dither (96/24 eats too much CPU time). Similar to the oversampling done with high end CD players. Verdict is still out, but overall seems to sound more relaxed, dynamic and "analog-like" with less grain, just what I was looking for

.

 

[blessingx]

FhG usually still wins over LAME in low bitrates, but LAME takes the mid and top. I use AAC now mostly, but when I use MP3 it's LAME for music, but FhG for audiobooks, speech, etc.

Thanks for your comments Chadbang. I'm a little surprised you've noticed such a large difference between 3.90 & 3.93, though I know 3.93 introduced some changes and 3.90 is still recommended. I've used 3.92 for nearly all my encodings.

One little tip, use the "--vcomment" tag at the end of your --alt-preset to write your settings into the comment field. Also did you notice any difference between using the "fast" tag or not (it may not be available for -api)?

EDIT: gpalmer it looks like he was using iTunes modified FhG. Don't know many specs about it, though it does go to 320 kbps and has an ABR option and 10 Hz filter.

 

[badmonkey]

Take a look here: http://www4.head-fi.org/forums/showt...threadid=50554 and here: http://www.airwindows.com/encoders/index.html

 

[chadbang]

Quote:

I'm a little surprised you've noticed such a large difference between 3.90 & 3.93, though I know 3.93 introduced some changes and 3.90 is still recommended.

No, Blessingx, it wasn't a big difference by any means. Very slight, yet I (believe) I did notice a difference.

gpalmer, if forget what version of the fraunhofer, sorry. I'm at work and I did it on my home computer so i can't check right now.

fewtch, that's interesting about pre-echo. thanks for the interesting info!

Blademonkey, nice article! Thanks. this one, folks: http://www.airwindows.com/encoders/index.html

He says LAME is a great encoder UP TO 256K, but falters at 320k! Very interesting news and may account for my first bad experience with LAME mp3s recorded at 320k. too bad he doesn't go into VBR. But now I will try some controlled bit rate LAME encodings at 256!

 

[Glassman]

Quote:

Originally posted by fewtch What you're probably noticing is less pre-echo with Lame (a particularly nasty MP3 artifact that smears transients -- in particular, instrument attacks). I was feeling a bit unhappy with how "digital" my soundcard has been sounding, so I went and found a hard-to-find plugin for WinAMP (SSRC resampling) and am listening to all my music files (compressed or uncompressed) upsampled to 88.2/24 with dither (96/24 eats too much CPU time). Similar to the oversampling done with high end CD players. Verdict is still out, but overall seems to sound more relaxed, dynamic and "analog-like" with less grain, just what I was looking for .

I'm always listening to foobar SSRC resampled to even 176.4kHz or 192kHz without dither and found it less harsh than plain 44.1kHz even though the DACs on my card do internally oversample everything to even 5.65MHz or 6.15MHz, but apparently SSRC does it better..

 

[lindrone]

No comments to add.. just wanted to say good job on a pretty painful set of comparisons... all those encoding and experimenting can get rather annoying..

Good job chadbang