Did you ABX this?!

No need to ABX try this sample http://www.volny.cz/jmartis/hissing_testsample.flac (sorry it comes from a lossy source) and encode with lame V5 the hissing it pretty evident!

Edit- I just realised FHG produces the hissing too! I must have missed this previously, I'll probably have to investigate more before I make another statement
There's still one reason why I would use 128k cbr over Lame V5 and it's because lame tends to produce larger files usually around 140kbps with my music.

If average bitrates over 128 are that much of a problem to you, why don't you encode with LAME -V6 --vbr-new and ABX against Fraunhofer (or LAME for that matter) CBR 128? I think you'll be pleasantly surprised.

Edit: imho, a bit weird for someone with a 297 posts record, most of which in the MP3 - General section, to still come up with stuff like this.

Well I need to have the MP3 at 128kbps CBR to upload to soundclick, otherwise they'll automatically transcode to 128kbps CBR which sucks.

But I won't use CBR for any other reason. Not the original poster's reason, either.

I'm afraid a real answer isn't possible whether or not Lame is better than FhG, and the common beleive that Lame is the winner may be right but cannot easiliy be proven.
IMO due to experience in the 192 kbps range

• Lame 3.98's VBR implementation is superior to FhG's.
• Lame 3.98's pre-echo behavior is superior to FhG's
• FhG has some usage restrictions. Joint stereo is available only up to 192 kbps. Seems like the FhG people don't have a lot of confidence in their joint-stereo implementation and prefer plain L/S stereo with very high bitrate. This restriction probably doesn't effect the larger part of mp3 users.
• As a tendency tonal problems are more severe with Lame up to 3.97 than is the case with FhG. Things have changed with 3.98, and Lame 3.98 -V2 in an overall sense (with the samples I tried) is pretty much on par with FhG now.

Because I'm pretty insensitive to pre-echo but find tonal problems pretty annoying I used FhG CBR 192 (the version that ships with dbpoweramp) for quite a while with my DAP and was very happy with the results.
I would have preferred using CBR 224 for safety reasons but as FHG uses plain stereo in this case I preferred CBR 192. Didn't run upon any stereo problems. Don't care about CBR at very high bitrate (did care at low bitrate but meanwhile I'm not even sure whether VBR at 128 kbps is a big plus. Most members beleive it cause due to theory it should be - and at 128 kbps I still think so too - but a real comparison may be very interesting).
Propably I'd prefer current Lame 3.98 now but I'm not sure.

Anyway things depend on personal needs.

My experience arises mainly from problem samples at >~ 190 kbps.

shadowking's proposal of comparing regular music at 128 kbps would be very helpful (Lame most current 3.98 -V5 vs. FhG CBR 128).

True VBR is overrated. Above 128k I prefer ABR (safe vbr). I found out if 128k sounds a bit crappy then --abr 144 usually sounds good - like V5 and bitrate is similar. I went back several versions with simialr results then I took V3.89 (pre alt preset era) built and still good results at --abr 144. I used gogo abr 160 k for a long time without problems too.

Several weeks ago I found weird issues like new vbr vs old vbr, some issues like vbr worse quality when psymodel fails etc. I am out of the vbr game. For hifi I use --abr 280 and --abr 144 for portable.

Not saying true vbr is worse but it seems like there was too much politics - r3mix vs APS vs MPC and everyone was out to find perfection preset to out do CBR 192. From what I gather reading far back is that 128k cbr sucked and 192 cbr still wasn't good and this started the war of presets commandlines etc. 128k / 192k were 'out 'infavour of other formats . IMO --abr 144 would have solved the 128k dillema and --abr 192~256 the high bitrate dillema. Also I read that LAME was chaotic before 3.90 but didn't find any evidence of this.

Even many years ago LAME would have taken on MPC, vorbis and 128k if --abr 144 was used:

http://www.rjamorim.com/test/128extension/results.html

So true VBR is NOT an advantage in FHG or another non-Lame encoder unless the implementation is very good and the psymodel is high-tech stuff (its not with mp3).

Yes, at least with low to moderately high bitrate ABR should be preferred over CBR. The availability of ABR is a great plus for Lame I forgot to mention.

Maybe, but it's not certain. And in all cases this boosted LAME setting would be clearly unfair: --abr 144 implies an average bitrate of 140...145 kbps on a complete MP3 library whereas all other competitors of the past test got their settings carefully scaled to not exceed 130 kbps on a wide audio set.

___
But I share some parts of your message. I also wondered in the past about the similarities between -V4 and --abr 160. Here is what I wrote 30 months ago:
< source >

But note that I still don't have any certitude on this question. ABR may appear as safer and stronger than VBR as long as we don't find it's own Achilles’ heel.
But in any case VBR has one undeniable advantage ABR doesn't and won't have: efficiency. ABR will give you the same kind of bitrate whatever your music is. VBR don't - and your bitrate should drastically drop in several situation:
• monophonic recordings
• quiet music (piano,...)
• music with few sf21 issue (ex: most classical music recordings)

As a consequence, even if we can say that ABR offers the same quality level than VBR or even more security (I repeat: I simply agree to call in question the taboo of VBR's superiority - nothing more) it loses on the efficiency side.

I wonder who's trolling here.

The question asked was about Fraunhofer versus Lame. The sticky posts do not talk about Fraunhofer, so it was a completely off topic answer.
In the rest of the forum, the only answer we can find is "Lame is more tested than FhG", which does not answer at all the question asked.



This test is more than three years old. Both Lame and Fraunhofer have evolved since then, and moreover, Roberto Amorim states that in that test, it was a mistake to use Fraunhofer in VBR, because it was tuned for CBR.

So in order to answer the question, get the latest stable release of both encoders, and start ABC/HR !

Note : Jimpin was banned for another reason.

my main use of fhg is when encoding xvid films which need cbr encoding
lame lacks cbr at 96kbs and needs 112 and above

I am willing to perform an ABC/HR test between Lame and Fraunhofer. I would perform ABX tests on most samples in order to give meaningful results.

My idea is to use the 20 samples of the Sebastian Mares 128 kbps multiformat test. Lame gave a 144 kbps average bitrate with them, however, I've just encoded them again, and I get 139 kbps. I guess that Sebastian's samples were not encoded as such, but taken from the whole track encoding.

Before this, I tried the mp3sencoder from Fraunhofer, and saw that Lame was between -m 3 and -m 4. Lame is also said to give a 130 kbps average bitrate with -V5, but I didn't find the statistics from which this figure is taken.
Since Fraunhofer was said to be tuned for CBR, CBR 130 kbps and CBR 139, or CBR 144 kbps should also be included.

I was going to start the test, but checking the files, I saw that they were clipping, and that they were offsetted. I'm too lazy to perform all the necessary adjustments.

So if someone else is interested in such a test and willing to deal with scaling and offsetting the samples, we may proceed. Otherwise, I don't think that I will spend the time to do it.

LAME dev mail archive: LAME 3.98b3 -V5 detailed result
LAME dev mail archive: LAME 3.98b3 all compression levels
LAME dev mail archive: used files list as attachment

Why not use "-b 96 --resample 48 -t"? (However, I never ever use it.)



I think java ABC/HR can handle anything annoying like that. But latest FhG enc/decoder supports gapless decoding (-ofl switch). Clipping caused by lossy encoding isn't audible most of the time and we have to live with that (unless he/she is Replaygain user).

Thanks for the info.
As we can see in the third post, the samples used are pop-rock from the 70-80's (Kraftwerk, Pink Floyd, Eurythmics...). That is quite dynamic pop-rock.



Thanks for the decoder tip. I don't see how ABC/HR would detect offset.
I use the W32 binary 1.1 beta 2, because Java's audio API is not very reliable. As far as I remember, it may click starting the playback of some files.



Since clipping is not related to the encoding performance, and can be avoided with replaygain, playback volume, or scaling before encoding, I prefer testing without clipping.

That can be avoided with latest build and listening test preparation (discard first 2sec of samples). I rarely see serious trouble about it. I'd recommed same method as public listening test, which is not well documented.

The new FHG is very neat . I encoded few tracks and even 128k cbr was decent. This cbr/freeformat mode is very interesting and I think its actually better than vbr for lower bitrates. I used 140 k cbr and it sounds great. The speed is amazing and I only used the fast mode. I would not be surprised if this encoder beats LAME..

MP3 developement is astounding. Now we have three good encoders: LAME, Helix/Xing and FHG.

I'm just curious, is joint stereo very important at 256 kbps and 320 kbps? Its critical to save bits at lower levels but does it matter at high bit rates? I'm curious because I'm anticipating Fhg 256 kbps MP3s for sale from EMI on Amazon soon. That's my best guess as to the encoder and bit rate - to match iTunes

For the record - I use LAME 3.97 -v 0 for my encodes. However, I cannot really ABX between this and Fhg 256 kbps. Others with better ears may be able to.

There's not gonna be any difference at 256 on most stuff- at least with my experience. LAME actually uses less MS frames a very high ABR / CBR - 240~320k. At this bitrate the only problems you get is mp3 limitation in preecho - EIG, fatboy, highhat etc.

Good news 256 kbps FhG mp3's will be buyable.

My remarks were aiming at own encodings, and as I was happy with CBR 192 joint stereo there was no reason for going CBR 256 plain stereo.
But the other way around there is no reason for not using a given CBR 256 plain stereo FhG encoding. Quality will be great (except for pre-echo which is a problem in general but towards which not everbody is very sensitive [I'm luckily not]).

Is there any decoder except the one they provide that is able to take the "OFL" data into account?

Link (from the changelog)

The next version of Winamp will.

Good news. Thank you.

I'm ripping and encoding a cd right now using audiograbber (don't ask me why) and the latest mp3sencoder (v1.4) at the highest vbr setting, with original file length (-br 0 -vbri -ofl -q 1 -m 1).

Aside from the earlier post that it somewhat sounds like pressing a typerwiter, what are your opinions regarding this latest FhG encoder?

My observations:

• So far it's free to download and use (may have patent issues, but they still had it floating in cyberspace)
• It appears to be a glorified fastencc (stereo bug still?)
• Amazingly fast
• Simple stereo at its highest vbr setting (this would be unacceptable to others)
• Sounds good to me though

Aside from the proposed 128k head-to-head comparison between LAME and FhG, what are your thoughts on the VBR (at least at high bitrates @ the same frequency cut-off, like APS and an FhG VBR @ 19khz)?

I wouldn't agree to the LAME VBR ABR 144 though, must be done both at forced CBR 128.

EDIT: wrong mode...

What about foobar2000?

Hello to this nice community! I'm Italian, sorry for my bad english.

I'm following the development of the Lame encoder from about 5 years, and I think it is great.
But... i encoded for my car audio Mp3 reader with L3.98b4 V0 one of my well known preferred albums
(Bob Dylan - At Budokan) and, nothing wrong, but feeling something like poor (poor life, poor involvent).

Then I encoded the same album with foobar using Fraunhofer mp3sEnc v1.4 (line:-if %s -of %d -br 192000 -q 1 -ofl) and i found it very, very good in my car. I found it very good also in my earphones, not evidently,
but easily abxed. I seems (I'm sure) Fraunhofer development is dramatically improved.

I Know, I'm telling something like blasphemous, but i want to know the truth (i don't think it is subjective)

Ciao!

You abxed it using foobar's abx tool?
Can you tell a bit more about it: with which track at what spot is it most easily abxable to you, what's your score when abxing it?

(I ask cause it's a bit similar to what I found using Lame 3.90.3 which was my favorite mp3 encoder some time ago. But I did not get very reliable repeatedly good abx scores so I wonder whether that's placebo. FhG however - I'm used to the dbpowerAmp version but don't think there will be big differences qualitywise with the various current FhG encoders - has never brought any doubt to me @ CBR 192 as far as tonal quality is concerned.)

Ok, halb27, thank you.

Yes, no one is immune from placebo, maybe me too in this issue..
In fact (damn..) this, morning (to give more details) i tried
some comparisons between two tracks encoded with Fraunhofer 192
and Lame 3.98b4 -v0 with no acceptable score.
Not enought time may be, sometimes my wife was bothering me, and now
I must run to work.

I don't use Foobar abx tool, but my own method...
I don't know dbpowerAmp version, only the olds slow encoders, the fast encoder
and the Acm v3.4.0.b0: Always i found them good, but not like Lame.

I'll try another time in some days.

Ciao!

Own methods are suspected to be not valid.

Outside of this forum things are easier.
It is sufficient to feel a specific encoder is best in order to use it with pleasure. We (most of the time) should stick to our emotions. It's a proven superior decision making strategy against nitpickingly weighing arguments. Anyway this is about personal preference.

Here however we should learn about valid abx results.

IMHO Fraunhofer have better low bitrate encoding <96kbs, it have Intesivity Stereo and Narrowing Stereo Image tool for very low bitrates.

Also i agree with Fraunhofer policy for Mid/Side JS coding, realy i know that is process totality lossless, but realy i'am feel safer when disable it on very high bitrates

Yes on the "Fraunhofer has Intensity stereo" and that this process helps getting more quality on the lower bitrate range. Not proven onto the "better low bitrate encoding".
I have a feeling it is more like a tie between both at their current stages, with frauhoffer *maybe* having a small edge over the other.



If i gave you 100 bucks for each 24kbps mono mp3 file you encoded, maybe you'd feel better encoding at that range.

Seriously, feelings are sometimes wrong. That's why we test things here. There have been a few cases of joint stereo being worse than simple stereo, but most of the times that was due to deficiencies in the way that was tested ( isolation of channels that don't represent the whole thing or simply ignoring that the better stereo was actually generating other artifacts).
The amount of bitrate is limited in MP3, and the more you want to store, the higher the bitrate needs to be.

I would recently have suspected the same thing, given that LAME doesn't even try to implement Intensity Stereo at low bitrates, which might seem a sensible tradeoff to reduce ringing and flanging type artifacts by throwing more bits at the mid channel.

However, in the MP3 encoder selection pre-test (in FRENCH) prior to Guruboolez's 96 kbps multi-format listening test (also in French), LAME came out better than either of the Fraunhofer encoders tested by a statistically significant margin (>95% certainty for mixed group, slightly less than 95% for the non-classical group). This was 2 years ago and according to Guruboolez's excellent hearing not yours, but with sound double-blind testing. Both encoders have changed a little since, but it's a large performance gap for FhG to make up.

I haven't read any tests actually below 96 kbps to compare a relatively recent LAME against Fraunhofer.

N.B. You can view Google's Automated Translation of the whole thread into English but you should note a couple of important mis-translations:

• English "Hen 1" - should be "Pool 1" etc. from french poule, which can be translated as the aforementioned poultry bird
• "BLADE" should read "LAME" most of the time , with wonderful irony uniting the best and worst MP3 encoders. The french noun lame means the blade of a knife, but blade is not a French word so comes out exactly the same. As the words in graphs are graphical, they're neither translated nor mistranslated
• "flow" should in this context have been translated as "bitrate"

The translation stops and returns to French someway down the thread after the MP3 selection, but it may be possible to view posts individually or simply paste the French into Google.com/translate.

It is a very informative and well-conducted test that is great amunition against people who claim that WMA at 64 kbps equals MP3 at 128 kbps, parrotting Microsoft's marketing claims from some time ago. Good LAME MP3s at 96 kbps actually beat the best WMA at 96 kbps, let alone 64 kbps, and 128 kbps LAME ABR was used as the high anchor and is far ahead of WMA96. Vorbis AoTuV at 96kbps seemed at the time to be the first to beat the current best MP3 at 128 in double-blind testing at lower bitrate, though not on classical (where it did well), just on 'varied' music (where it was exceptionally good). These comparisons cannot simply be read across to comparisons against current LAME -V5 (~130 kbps) or LAME -V7 (~100 kbps), which differ from the tested LAME settings in non-trivial ways.

by the way is LAME 3.98 beta 4 stable ? , will it give me the same results of 3.97 ?

now i'm using 3.97

a) 3.98b4 is stable.

b) 3.98b4 will not give the same results as 3.97.

Regarding quality there is no final answer. As for my best knowledge in an overall sense

- for bitrates ~ 128 kbps the results that are published so far seem to be in favor of 3.97

- 3.98's VBR mode seems to be more stable. So at a higher bitrate with a setting like V2 or V1 3.98 seems to be preferable.
This is especially true for certain tonal problems like trumpet or Birds with which 3.97's quality is worse than that of other encoders.

- For best quality like ABR ~ 270, CBR 256 or CBR 320 both versions are of the same good quality in a pratical sense.

This question has to make wanting to commit suicide, (or homicide) to any developer.

"Hey! I see you're still painting my house, but you've finished, right?"

beta is beta. And nothing is stable.

Personally I like Fraunhoffer IIS for encoding 112k and below. I am not crazy about Lame, too much metallic hiss in higher frequencies. I prefer older mp3 codecs, usually 224~256k for solid mp3 sound. (APE is my fav non mp3 codec).

Wow, what is it with this thread that attracts the flamebaits like ants to a picnic?

Huh??? hiss as an mp3 artifact?? This is something new to me.

"Wow, what is it with this thread that attracts the flamebaits like ants to a picnic?"

The same thing that attracts the LAME fanboi's to the picnic.

Perhaps MisterMeow would like to post some examples encoded with both encoders for others to see, or risk a TOS #8. Presumably at this bitrate asking for ABX results is futile.

What's better, FhG or LAME?
Answers:
From (censored)@iis.fraunhofer.de:
Of course we, as Fraunhofer Institute only recommend our Fraunhofer MP3
encoder.

Regards
Mandy Trommler

From (censored)@codingtechnologies.com:
I would suggest to use the Fraunhofer MP3 Encoder, but better use aacPlus
instead of MP3 :-)

Christian


Whats the latest version of Fraunhofer MP3 Encoder & where i can download it?

Answer from (censored)@iis.fraunhofer.de:
No problem! You can find our latest command-line codecs (encoder version 1.4, decoder version 1.3) on

http://www.all4mp3.com/tools/sw_fhg_cl.html

Silver Wave: May I ask you what should we understand from your post?

Fraunhofer answer is clear enough to need comments

Coding technologies is nowhere a neutral party, and it also tells to use its technology intead of one of others.

And i guess we are aware of the mp3 surround encoded (Which was pointed previously in this same thread: http://www.hydrogenaudio.org/forums/index....5784&st=25# )

Guess we cannot get a clear answer, because:
a) there's a too wide range of situations where to use an encoder (low bitrate cbr, portable bitrate cbr/vbr, high bitrate cbr, transparency level)
b) there's not enough sample values on the tests done here at hydrogenaudio to get a unquestionable answer, even when they are a good guideline on what to expect.
c) being two completely independend projects, both have strong and weak points in different places each. Only an average of the results can have a meaning as a whole, and that requires many samples.

Anyway, i guess the best we can do here, is wait for (and participate in) the next mp3 at 128kbps listening test, that is scheduled after the one that has just finished.

Fraunhofer's advantage is speed. It is significantly faster than LAME, at least on the Pentium IV and the PowerPC where there has been very good optimization done.

FhG MP3 Surround Encoder (specially) optimized for Pentium CPU's.

Has any testing been done on mp3sEnc's VBR modes?

Not much with mp3sEnc but quite a lot with other FhG versions. From that my experience is to better use CBR. (My experience was in the 192+ kbps CBR range).

Not much but some impulse samples.
Strong artefacts even in highest vbr quality. cbr is much better.

Edit: Try eig.wav this should be enough

Test both encoder in the 128kbit CBR Joint Stereo.
LAME 3.98 Beta 5:
http://rarewares.poskolio.com/lame3.98b5.zip

FhG MP3 Surround 1.4:
http://www.all4mp3.com/dev/download.aspx?n...14_20070711.zip

Just test the latest...

Thanks for pointing me at that sample, mp3sEnc VBR doesn't do much good there, even at it's highest setting "-m 1" artifacts are apparent from the start, 192k CBR gives better results. I also tried this sample with LAME -V 3, MPC --standard, and Vorbis -q 4, all gave noticeably better results.


With eig.wav at 128k, both produce annoying artifacts, mp3sEnc producing horrible clicks, I would choose LAME here simply because I find it less annoying.

Just a reminder:
Please don´t use lame cbr under any circumstances until you have to use it on incompatible hardware.
When talking about a given bitrate please use lame at a correspondending vbr quality or at least in abr mode.

Does LAME support surround?