I wanted to find out what the differences were between the various programs based on the Fraunhofer code. I also wanted to find out which version was the Newest. This comparison includes the version found in the newest version of Cool Edit Pro. You can find it here:

http://pws.prserv.net/spankdaddy/mp3/

I tried to make it easy to understand as I needed to understand it as well.

but why do you analyze spectral graphs ? they are nice to look at, but don't tell you a thing about quality



heh ? the best fraunhofer encoders yes, but not the best mp3 encoders... and for what bitrate ?

1. I definitely do not try to tell you which sounds the best, as I am not you. If you read the page I state that it is meant to compare the different encoders and show their differences, not tell you which sounds the best.
2. Hey. that was radium talking not me, it was a quote saying what encoder was used, not me telling you they are the best.

2. ic, sorry for that

I found that there were several families of straight Fraunhofer mp3 codecs (not talking about mp3pro here):

1. The Audioactive / Producer Pro / Radium family
2. The Mp3Enc / "Alternate" (slow) family
3. The Fast / VBR family

There may be slight differences from program to program within each family, but in general there are the same defining characteristics within each family. There were a couple of Fraunhofer releases which were buggy -- i.e., the Fast Codec "stereo collapse" release, and the Slow Codec "sounds horrible at high frequencies" release.

I don't know where l3enc falls; I haven't really looked at it, but if I had to take a wild guess, I'd say it belongs in the Mp3Enc family. There was also another Fraunhofer codec which I believe produced watermarked files. This was in Liquifier Pro 5.0.

The defining characteristic of the Mp3Enc family is that it is extremely slow when all the options are enabled. It also has a defining audible quality -- i.e., the low frequency glitching. The Fast codec is, of course, extremely fast. Fraunhofer VBR is based on the fast codec.

A bit of confusion to avoid: The "FastEnc" library used to come packed with both the fast and the slow codecs, but in Cool Edit Pro 2.0, the slow codec appears to have been eliminated. I typically refer to "FastEnc" interchangeably with the Fast codec, not the slow one.

I haven't looked at what Cool Edit Pro 2.1 has (I have looked at version 2.0), nor have I looked at the latest MMJB version, but I don't have any reason to believe that they would be much improved over previous versions. Does anybody have information to the contrary?

Confusing, ain't it?

ff123

Yeah the "Fastenc" is included in Cool Edit Prov2.1. It definitely produces different files than the HQ mode. And stuff has changed like the fixed dropout bug. Take alook at it http://pws.prserv.net/spankdaddy/mp3/

The Radium / Kristal Studios codecs are not the same thing. At certain bitrates / quality settings / channel modes they produce different results. Set them both to 96kbps / low quality / MS/IS to see what I mean. Further testing beyond 128kbps is needed before you can declare that "... it's exactly the same as radium, even though Kristal is dated 2001! All these guys did was change the info to read kristal." Their output is very similar, but not the same at many points, and the KS codec IS considered to be latter / slightly better that the Radium codec IIRC.

The Fast codec never had the dropout bug. The dropout bug happened only in the slow codec, which was dropped from Cool Edit Pro 2.0. So probably that codec isn't there for version 2.1 either.

ff123

What is the relationship of Kristal Studios to Opticom or Audioactive? Are they the same thing?

I've heard that Radium dropped one of the encoding options to speed up their version relative to Opticom/Audioactive, but I've never verified this. Also, it's true that I never investigated differences at low bitrates to see how IS differed in the various flavors.

ff123

Codecs listed are:
Fast Codec (High Quality)
Medium Codec (Average Quality)
High Quality Codec (Slowest)


btw, my original post to the Syntrillium forums that verified CoolEditPro2.1 having a new FhG codec has since been deleted, I never found out why or what was posted in the last few weeks, if anything. Just thought it's kinda interesting they deleted the whole thread instead of just deleting the message in question.

now i really do want to test out the difference between 2.0 and 2.1 with Equal and see if the two HQ codecs are really different.

You'd definitely know if Cool Edit Pro 2.1 was using the real slow codec. I'd say it
encodes at least 10 times slower than the fast codec. BTW, I believe the categories above are the same as for Cool Edit Pro 2.0.

Nope it never did, wrong wording, both the fast and the codec previously referred to as alternate have changed in cool edit pro 2.1 though. The dropout bug has been fixed though and the general output is different.,


Ummm, I just tried it and they turned up exactly the same thing. why don't you post your results and how you got them.

What I meant to also say is that the "Alternate" or slow codec is no longer included in Cool Edit Pro 2.0, and probably also 2.1. It's all just the Fast codec now, probably with various switches turned off to make it even faster for the extremely fast option. That's why there is no dropout bug -- there is no codec with this bug to "fix."

ff123

L3enc is a different familly.
It belongs to the v2 familly, and the successors of this familly are Mp3enc ones (they have been reffered in the past by FhG as L3enc v3)

It seems that Mp3Enc is an evolution from L3enc, while Audioactive/Acm/Opticom are another familly, not directly evolved from L3enc although they appeared before the Mp3Enc familly.

It would be interesting to compare FastEnc from CEP 1.2a (with mp3-me plug-in) and the various modes of the encoder in CEP 2.1 to see if any were the same.

Have you tried this ff123?

Cheers,
David.

From there:
http://web.archive.org/web/19980519144106/.../amm/index.html
we can see that Mp3enc and Producer were released at about the same time, but with different goals.

I'm just wondering. I'm using Proteron's EasyMP3 1.1 program. It has the fastenc engine and alternative. Am I correct in assuming that the fast setting is recommended in this version? The high quality setting produces those high artifacts problems. Also what about EasyMP3 1.0 is the fast codec or alternative recommended in there? just curious, because I'm a little confused.

No I haven't. I compared with CEP 2.0, though.

I believe I found that CEP 1.2a with MP3-ME used the fast codec for the fastest setting, the slow codec with lower quality for the medium setting, and the slow codec at full quality for the best setting. The slow codec at full quality was the buggy one (and maybe the medium setting was buggy too).

CEP 2.0 used the fast codec exclusively, the only difference between the settings being things like lowpass (no lowpassing for the fastest) and perhaps some quality setting.

I don't have CEP 2.1 installed on my computer. Actually, currently no version of CEP is installed on my computer.

ff123

Don't use the highest quality setting in EasyMP3 1.1. That's the buggy slow codec. I don't know about version 1.0.

http://ff123.net/notrecomm.html

ff123

I think what your saying is that all of the settings in cool edit pro like
Fast Codec (High Quality)
Medium Codec (Average Quality)
High Quality Codec (Slowest)
Are all based on "fastenc"
If so the help file says something different:

The fast option in both Cool Edit 2000 and Cool Edit Pro 2.1 both output files that resemble each other very closely. (see the Comparison) However the High quality setting in Cool edit pro 2.1 is noticebly different than the fast option in either program and the alternate codec. So if it is based on fastenc it is the high quality version that was tweaked up. After examining some more files I do see that perhaps the High quality mode in Cool Edit Pro is related to fastenc but it is very different than prevous versions.

You should be able to perform a listening test to figure out if the high quality codec is related to the Fast codec: try main_theme.wav, which can be found here:

http://www.mp3dev.org/mp3/gpsycho/quality.html

at 128 kbit/s

Problems with this sample should be quite noticeable with the Fast codec.

ff123

Edit: When I re-install Cool Edit Pro 2.0, I can try this using that version
Edit 2: But I think I have already done this at the highest quality setting. It would just be to verify that the fast setting has the same problem.

ermm i'm just gonna contribute this quick snip of info.. a while ago on the wavelab support forums, the developer was asked a question about the fhg encoder used in wavelab, he said that fhg made quite a few "variants" of their latest encoder, fastenc, and that wavelab was using one called the "pro" version. he also said that one of the reasons this encoder was dropped and switched with LAME was the expensive licensing fee for this particular codec. i think (and i re-emphasize think) fastenc is still available on wavelab, but you will have to pay an extra licensing fee. otherwise the program defaults to LAME

I don't get it. Whats the point of comparing second rate codecs?

In case all the modern codecs start to sound too good?

I'd say lame and FhG's fast codec are roughly equals at 128 kbit/s, so from this standpoint we're not talking about "second rate" codecs.

ff123

I just now did a listening test of CEP 2.1 compared with the slow codec in fastencc
1.01 and lame 3.90.2 using --alt-preset cbr 128 --scale 1. I used main_theme.wav.

The two CEP codecs were 1) the highest quality setting (selected by default when using 128 kbit/s) and 2) the lowest, fastest setting. The CEP codecs sounded pretty much the same to me, and were far worse than than the FhG slow codec and lame. FhG slow and lame were not significantly different-sounding to me; maybe just a slight bit noisier for the FhG slow codec compared with lame.

The CEP codecs sounded watery. I think they both sound pretty much the same on this sample as they did in CEP 2.0. However, it must be said that I'm going on memory for the sound of the CEP 2.0 encode.



ABC/HR Version 0.9b, 30 August 2002
Testname:

1R = D:\junk\main_theme_fast.wav
2L = D:\junk\main_theme_slow.wav
3L = D:\junk\main_theme_highest.wav
4L = D:\junk\main_theme_lame.wav

---------------------------------------
General Comments:

---------------------------------------

1R File: D:\junk\main_theme_fast.wav
1R Rating: 3.6
1R Comment:
---------------------------------------
2L File: D:\junk\main_theme_slow.wav
2L Rating: 4.6
2L Comment:
---------------------------------------
3L File: D:\junk\main_theme_highest.wav
3L Rating: 3.6
3L Comment:
---------------------------------------
4L File: D:\junk\main_theme_lame.wav
4L Rating: 4.8
4L Comment:
---------------------------------------
ABX Results:

I've actually found my magical sweet spot to be Fraunhofer FastEnc (from cep2) with a 90-115kbps vbr. I like it so much that I even opened my mouth to speak of a new lowbitrate standard, only to have several people promptly vomit upon me. I'm aware of minor imperfections but honestly don't really mind. When I get my state of the art sound system I'm sure I'll have some spare cash to be buying CDs.

One more thing concerning main_theme.wav

If you look at the spectral view of each encode, it seems very clear that the two CEP 2.1 settings are in the same codec family. The spectral view of the slow codec in fastencc.exe is very different from these two. The lame spectral view is different from both fastencc.exe and the two CEP 2.1 spectral views.

ff123

OK GUYS LISTEN UP

The FhG Codeds aren't perfect but they are high quality codecs with a valid place. If you compare the lower bit rate work from LAME I think you may find you prefer the FhG output (you may not, however my ears do!).

Moreover take a trip to the LAME Project Page http://lame.sourceforge.net/ and read it's description of LAME:

" Quality is comparable to FhG encoding engines and substantially better than most other encoders."

If you are one of these guys who spends his life flaming FhG users print this off and put it by your monitor as a reminder that people who know LAME better than you or I respect the strengths of FhG.

Now that said.... isn't APS sweet ;-)

Right must get back to ripping my copy protected CDs....

Fair Use and Love,
Fairy

Oh and some of you may be able to help me too ;-)

http://www.hydrogenaudio.org/forums/index....ST&f=16&t=15589

Heh half the stuff on that site is ancient. I'm sure that was true at one point, probably the late 90s

I think even now FhG slightly edges out LAME at bitrates under 128 kbps.

LAME v3.93.1 December 1 2002:

* Many improvements in quality in speed. See history .
* MPEG1,2 and 2.5 layer III encoding.
* CBR (constant bitrate) and two types of variable bitrate, VBR and ABR.
* Encoding engine can be compiled as a shared library (Linux/UNIX), DLL or ACM codec (Windows)
* free format encoding and decoding
* GPSYCHO: a GPL'd psycho acoustic and noise shaping model.
* Powerfull and easy to use presets
* Quality is comparable to FhG encoding engines and substantially better than most other encoders.
* Fast! Encodes faster than real time on a PII 266 at highest quality mode.
* MP3x: a GTK/X-Window MP3 frame analyzer for both .mp3 and unencoded audio files.


No dude it's current and it is THE LAME Project Page. It's THE source from the guys who make LAME happen. Personally I reckon they know a thing or two about codecs, moreover my listening testing has shown (me) that LAME excels with the right settings (e.g. aps) but FhG is also very good especially at lower bit rates, even more so if you tweak the settings to mimick what people have found they like on LAME.

Your response is exactly the sort of closed minded approach that is worryingly common but far from universal.

Do some listening tests below 192kbps FhG rocks, at 192 and beyond LAME tends to do better. LAME aps is sublime!

I do a lot of portable listening and am trying to find a good alternative to APS with a slightly smaller file size.

Fairy

Calm it down for me, baby.

Why so much agressiveness? It'll be bad for your ovaries.


Anyway: yes. FhG is better than Lame in two circumstances:
-Low bitrates: Because lame was always targeted at transparency and IS was never implemented
-Frequencies other than 44.100Hz. Because that's the CD frequency, and the coders didn't care much about other frequencies - they had few time, so it would be better used tweaking the most important frequency.

Regards;

Roberto.

You're right I was getting on my high horse, sorry all. It just gets like a religion this LAME business people get slapped down for using FhG or called heritics for the crime of experimenting with the sound produced by different switches instead of just blindly following APS (which, did I mention, is great!), I reckon there's a danger of LAME being stifled by people not recognising and working on it's weaknesses.

It isn't perfect but it is very good.

Anyway I must get back to my ABR versus aps -Y considerations (apm sounds rotten to my ears!)....

Fairy

According to guruboolez, Fraunhoffer does better than LAME with harpsichord-like sounds:

http://www.hydrogenaudio.org/forums/index....ndpost&p=151564

Do these tests / results apply to MusicMatch too? After all it does use the Fhg encoder, which makes it one of the most widely used encoders around. Esp. since it's a free mp3 encoder in a free program.

After all, it too has three levels of encoding levels. They call it 'cpu level' of normal, high and very high, but that seems to compare to the three types in your other packages.

If you test MusicMatch, you might want to do one of the 7.x versions (7.1, 7.2 or 7.5) and either 8.0 or 8.1 (8.1 is particularly bloated, jumping from 10m to 17m)

Many people still use the older v7.x versions since they aren't quite as bloated and don't have the same online registration issue that v8.x does. (And the 8.x upgrade really caused problems for a lot of people who purchased 7.x expecting to be entitled to a free upgrade (or lifetime upgrade) to 8.x, like the website & box said. So a lot of people are staying with 7.x, so it might be worth testing as well as the current 8.x version.)

Just because the change-log is updated doesn't mean that everything is up-to-date. That list you refer to have been there a long time and hasn't changed at all for as long as I can remember. For instance, they refer to gpsycho as the psymodel used in Lame. But all currently recommended presets use the improved nspsytune psymodel, which gives better quality. The statement "Quality is comparable to FhG" is probably correct with the deafult settings and gpsycho, but not if you use the presets. As Mike said, "half is ancient", not everything.

Nevertheless, FhG is better at low bitrates and other sampling frequencies than 44.1, as rjamorim already stated.

Err type "Quality is comparable to FhG encoding" into google. Note that previous builds of LAME also claim that.

I suspect that if the pages still existed, every build in the last 3 years would claim that as well.

does something cease to be true because it is an old truth?

Thankfully we are mostly of the opinion that FhG is good for low bit rates and has a dreadful VBR header problem in many implementations (Musicmatch for one) but that Lame tends to hold the higher bit rate ground well.

What is your own view on the matter?

Love and warm hugs,
Fairy

Well, how about this analogy:
A couple of hundred years ago, the "truth" was that it was absolutely impossible for humans to fly. But due to technology development, that is no longer the "truth".

Same thing with Lame really. It has been developed a lot since that statement was made.

Now it's not so important, but where is (at least) 2 versions of MP3 ME! plugin for Cool Edit Pro with real "Alternate"(slow) codec.

1) 1999©, file date Feb 16 01:00:00 2000, link date Feb 14 22:21:28 2000
2) 2000©, file date Jul 20 01:00:00 2000, link date Jul 01 03:50:52 2000

Version 1) have (in terms http://www.ff123.net) "Alternate-bad" codec, but version 2) have "Alternate"(good) codec - at least on 128 and 192 in high mode with default lowpass it produce same output as fastencc 1.02 with -hq switch.

If you are like me, and don't believe everything you read?? If you prefer to judge by what you hear - then follow my suggestion, and get yourself a copy of "FRAUNHOFER Opticom or Audioactive Professional" Pro MP3 Encoder and then do some listening tests. (Fhg RADIUM Codec '99 produces much the same result)

Thats what I did, and the results were not only unexpected, but shocking!!

I encoded 12 different wav samples of selected acoustic material. I use samples that have an empasis on piano, strings, brass, cymbals, even a choir. If you know what these instruments sound like in real life, its easier to determine if they are being TRUTHFULLY encoded. I prefer to judge with audiophile recordings of classical or jazz, as this music generally has no overdub, or over production, and careful microphone placement.

Result? How did the encoder from 1998/99 do against Lame's latest? In each and every sample I worked with, the FhG-Opticom/AudioActive encoder exceeded Lame - in its ability to reproduce the sound of these real instruments faithfully. Clarity, definition, openness, realizm. I encoded Telarc Audiophile recordings to wave, then encoded at 128kbps using AudioActive Producer (the last slowenc codec from FhG to be tuned).

Let me repeat, I am only talking about the FhG encoder developed with Opticom/or Audioactive in 1998/99. I have also tested the FhG encoders previous to this, as well as those that branched out and became Mp3Enc and Fastenc (right up to Adobe Audition).

Lame sounds better at 128kbps than all the other Fraunhofer codecs, but NOT BETTER than the Professional Opticom or Audioactive/FhG encoders of 1998/99.

If you know what real instruments sounds like, then I encourage you to test these encoders - like me... you may be very surprised by what you CAN hear! Perhaps, as has been suggested to me by RJAMORIM - these FhG codecs have been worked on exhastively by technicians and listening experts who have fine tuned at 128kbps using mostly classical or acoustic music to do the tweeking.

Although Fraunhofer no longer is involved in working on improving these final codecs beyond any minor bug fixes, it should be stated that for encoding all general music types, Lame may still prove to be the overall winner. But if one is looking for a means of encoding the very rich detail of Classical or Jazz with impeccible honesty (at only 128kbps) - you won't want to over look Fraunhofer's most important codec for quality of sound reproduction. The fine tuning they accomplished over 6 years ago - still rules!

I now understand why the Lame project folks continue to suggest that FhG is the benchmark in sound quality for Mp3.