Well guys, after doing a lot of research over the past couple of weeks (ever since my Musicmatch plus 9.0 had died and I had to re-install something else *cough* iTunes *cough*), I wanted to find some things with regards to MusicMatch as it compares to other encoders such as LAME...in particular:

1. What version of the Fraunhoffer encoder does it use?
2. How good is the Encoding, particularly with VBR as compared to LAME?
3. What advantages / disadvantages does it have?

I contacted Musicmatch, and they would not give any answers with regards to exactly what version of the Fhr encoder they use, other than to say that they use several types of encoders, depending on the settings, and that how the Software decides is proprietary information. So I decided to find out by myself:

I used a CD from Jeff Lorber, and recorded it 3 different ways:

Using CDex 1.51:

Notice that the encoder used is LAME 3.92, and the setting I used was VBR 0 (max vbr I think). The majority of the frames were encoded in 192, with an average bit rate of 208.

Next, I used MusicMatch in VBR 98% (almost highest quality):

Here, you can see the encoder type was Fraunhoffer's FastEnc. The bitrate distribution was, in my opinion, much better distributed, and with a higher average (216). I like the fact that for the bulk of the song, he averaged around 224, but dropped down and raised up the bitrate as necessary (nice bell curve).

Finally, just to compare, I used MusicMatch in CBR at 256:

interestingly, the encoder used says "FhG (fastenc OR mp3enc)". I don't know if it means it used a combination of both, or it used mp3enc, or the software was not able to decipher between the two. But sufficeth to say, it is obvious that musicmatch does use different encoding when doing different types of recording settings.

Finally, I gave my friend's a "blind test" of sorts, and had them listen to the tracks (one was the LAME, one the MusicMatch in VBR) and tell me which one sounded better? 8 of 10 selected the MusicMatch . Here is the samples, and you guys tell me which one do you think sounds best:

lets see if you can tell them apart

Moderation: Links removed. Let's see if you can read the TOS.

Couple of other observations: MusicMatch recorded the song about 25% faster than LAME. The Size difference was not considerable between LAME's vbr and MM's VBR....but it was considerable when compared to MM's CBR at 256.

Why 3.92?

Id put on some flame retardent clothing if I were you

This test is very unreliable because only one track from one album is being examined.
Lames -V 0 to -V 5 vbr settings are designed to produce mp3s which sound identical to the track not which sound 'best'. At -V 0 it nearly always fools all high end testers on nearly all music types thrown at it, so if a test actualy did show that musicmatch is likely to be producing 'better sounding' mp3s than lames -V0, (or -V2 even) it would strongly suggest musicmatch is changing the sound of the tracks.
But because this test is so small its certainty of suggestion is very low and will not be regarded kindly here considering the forums Terms of service on dealing with useful data.

The mp3 tracks you are 'publishing' will also have to go if they are copyrighted.

Sorry mate, better luck'

Dunno....that's what came with CDex 1.51 when I installed it

yeah....I didn't know all that. sorry....just trying to share some info I found.

I can remove if the admins want me to

Considering the settings you used, I think that you might have better results with 3.97.
3.92 was released in early 2002.

Problem its not really info yet, just a tiny dislocated amount of data maybe that would have to be accompanied by much more compatible data before it started to become informative.

The Admins will remove the tracks if you dont first because theres a 30 second limit on samples here to protect the forum owners themselves from legal action.

You can update that in about 15 seconds by getting the newer lame.dll from rarewares.org and replacing the lame.dll in the CDex directory.

1) It doesn't use Fraunhoffer.
2) Its not good.
3) Offhand I can't think of any advantages. The disadvantages are that its a terrible old codec that shouldn't be used.



See the FAQ in this forum for proper LAME versions and settings.



See the FAQ at the top of General Audio for how to correctly perform a double blind test. Specifically, if you don't include the source file, the comparison is irrelevent.

1. Sorry Mike, but I don't know where you get your info....MM uses the Fraunhoffer codec. See the pics of the EncSpot details. You can also search the web or contact MM directly to confirm.
2. Not good compared to what? most people I've given samples of tracks recorded in MM vs LAME has preffered the MM ones....perhaps we are all wrong? LAME does make an overall better encoder I suppose, but in no way shape or form is MM not good.
3. What version of MM are you referring to as "old codec"? version 9.0 seems to be doing just fine with both the FastEnc and Mp3Enc of the FhG encoder...... Another very important advantage for me (since I rip music a lot) is speed, and MM showed at least a 25% speed advantage at worst case scenario...



yeah...I saw that. Thanks!! I upgraded it and am doing a rip right now to see how much better it is...


One thing guys...I did not post this thread to start a war or get flamed...I was here with the intent of sharing just some personal findings, that's all. I've been a long time reader (lurker I should say...hehehe) here, just never posted much, but I've learned a lot from you guys, so I decided to give a little back... If saying MM is a good medium of encoding music, is an insult to some who may think LAME is the best thing since sliced bread, then I apologize in advance....

I dont like to bring things to this level, but to be frank, this looks like trolling now, if youve been lurking here you should know better.
What you have given, is only an anecdote proposing a controversy to argue over and have explained at length so as not to mislead passing surfers into thinking the report sensibly suggests anything of worth about the relative high bitrate performance of lame and fraunhofer - its just way too small a sample to react to. Do you think if high bitrate performance was that easy to discern testers here would go to all the bother of doing hundreds and hundreds of controlled trials for each of their tests?

Keeping reading mate, id post links for you but I dont have them at hand and cant be bothered dredging them up, but rest assured the place is riddled with posters re-explaining statistical methods and testing requirements.

edit: hasslehoffer

From what I recall MusicMatch does use a couple different Fraunhofer engines. This is actually dependant upon the priority level you choose in the encoding/ripping settings.

A normal setting instructs MM to use the FastEnc encoder. I have heard that this is best use for bitrates above 128kbps (I think I read it on http://ff123.net)

If you select the high or very high priority it actually switches to an encoder that is very similar to MP3Enc. This encoder is better at lower bitrates.

The VBR is the same way. If you use normal mode it uses FastEnc, and I'm pretty sure if you change the priority that it will use the alternat encoder (similar to MP3Enc).

My mistake, I read iTunes in that sentence and missed MusicMatch. I should have been more careful. You should probably discard what I said above since I was comparing to iTunes MP3.

When will people learn that "Fraunhofer" is spelled with only two f, one in the beginning and one near the end?

You are just comparing both without including the originals?
I understand, sometimes an encoder, maybe FhG will add some ringing and high frequencies (hiss) in their encoded samples........ some people like that ringing/brightness effect??

In this forum, we are not talking about sounding good or bad, it's just whether an encoded sample is transparent or as close as possible to the original.
That's why we have ABX



if you want, you can simply search for LAME 3.97 alpha 8 (I don't know where to get it now), it's pretty fast, twice the speed of current LAME 3.97 Beta2 with "--preset fast standard", but I don't think you should use it!

It seems strange that in the Lame encode, there is not a single frame under 192kbps. It might be possible that under CDEx you selected the minimum frame size to be 192kbps.

The test is not blind, not using the original as a reference, but it might be possible that FhG brings better results than Lame considering the settings and version of Lame you used.
Using either the so called "alt-presets" or 3.97 would improve results.

I tried the latest LAME codec, and I still get similar results...the only difference was that the file size is a tad smaller now (.2 mb smaller). Sound quality is about the same in my ears

I did use 192 as the smallest setting on the encoder...is this wrong?

Yes, you need to use the -V switches. Use "-V 2 --vbr-new" in your sample, this will give you a VBR mp3 more matching the bitrates of the MusicMatch sample you took. You can look at the recomended Lame settings sticky at the top of the mp3general discussions.

This is why the encoder uses a lot of 192kbps frames, as you are preventing it to go lower.
I'd suggest to use 32kbps as a minimum, so the encoder is free to use the full range of frame sizes.

When I was encoding audiobooks a good while back, I felt i got better quality at lower sizes with MM/fraunhofer than LAME (can't remember version, about +2 years ago)
Havent done any newer compares in ages, since I started using higher bitrate and just stick with LAME for everything