Alpha 9 is now available.
Presets standard/standard1/standard3/standard4 are available.

Standard2 is gone, 0/1/3 are identical to alpha8. Standard4 is trying to compete against standard3.

Waiting for your results...

...Waiting for a compile

... gabriel.mp3-tech.org/lame

>_< Sorry... I searched on mp3-tech, but without succes.
I will post results later (this evening).

lame.exe and lame_enc.dll both available on my 'Others' page at Mirror 1 (link at bottom of main page).

You might want to redo that zip because it puts the exe and dll in a \lame-3.94alpha\output\ subdirectory upon extraction. Somebody might get confused.

Thanks, I will. I forgot about the path left over from a previous zip exercise!!

Edit: Just uploaded again.

I find it interesting that std4 is much faster on my P4 with generic optimizations rather than P4 specific ones.

You forgot to change the link name on your others.html page.
It still shows "lame 3.94 alpha 8 bundle" while is alpha 9
THX

I did some little test with std4 (not blind), on simple instrument.
· Erhu is buggy again (std3 was near perfection).
· Trumpets2 is crap
· French horn is crap (sample here)
· Tried on a Miles Davis sample, preset cbr 128 is better than std4

=> as a classical listner, i will throw away this preset, and I will chose ABR 150, much better.

Can someone confirm this impression ? I'm not enthusiastic about further testing with so poor initial results.


EDIT 1 : The horn sample on this page is buggy to (the same bug as erhu), when std3 is near perfect, if not perfect.
EDIT 2 : on the same page, the soprano sample is raucous too with std4, and not with std3
EDIT 3 : raucous sound with violin2 on the PCABX sample page, and not with std3

Ooops, sorry!! Just changed it.

Before I saw this new beta9 thread I was testing with Erhu but I think I'm pretty fatigued by it so I don't know if my scores are so useful anymore, but from my results Erhu isn't any worse with standard4 than with standard3. I'll mess with it later to try and confirm it though (that's the order of the tests as well):

Erhu:

3.90 APS 9/9 (Plop's there, no problem ABXing this one)

394 APS3 7/12 (pretty tough for me, had to really concentrated on noise rather than the "plop")

394 APS4 4/10 (headache and fatigued from this sample, I think the noise is still there in the high energy parts though I don't feel it's worse than APS3 at the moment)

Will test tonight.

OK, just encoded a couple Filter tracks with APS4 (You Walk Away & Where Do We Go From Here) and it's just REALLY bad. Obvious flanging. Something is clearly broke in APS4.

Well, at least it is interesting to know that std4 sounds bad...
It seems that until now, std3 is our winner.

Btw, I am not asking anyone to heavilly test std4, but if you have a track where it's really bad, it might help to post it.

I just moved this stream to the alpha9 cbr 128, using oddsock and dnas - coming from the dsp box winamp, so its a digital signal from the dsp...

http://66.129.124.253:8000/listen.pls

sounds pretty good to me, good pre-echo, stereo rivals Krystal Studio's Fraunhuffy 4.2



Let us know when the new compiles are up, and change the cbr 128 so it's the same as std3 if its the best then right?? And ill put it up when its out =)

jesseg

I tried the trumpets1 sample on all but the standard4 preset, and the small but audible high frequency artifacts are still there. I still need another round of abc/hr to know which ones sounds better or worse that the good'n'old --aps.

I'd imagine it would sound bad with most anything, but in this sample I find APS4 more offensive sounding than a --preset cbr 96 (along the same lines as guruboolez):

Filter - You Walk Away From Me (flac 1MB)

MP3 Encoded with --preset standard4 (260K)


Also with LAME 3.94a9 - preset standard3 I did a quick ABX which resulted in 8/10 based on slight flanging. With LAME 3.90.2 APS I got 8/10 again, but it seemed a better sounding than APS3.

I've uploaded again, this time to include the lameACM.acm, otherwise no real change.

Yes, according to my tests std3 is better than std4.

Sample: Trumpets1.wav
ABX Results:
std3 vs std4
12 out of 12, pval < 0.001
Comment: Obvious hollow noise, very easy to ABX

Sample: Liebestod.wav
ABX Results:
std3 vs std4
12 out of 12, pval < 0.001
Comment: Obvious hollow noise, very easy to ABX

I think this two samples are affected by the same problem, a strange hollow background noise.

I tryed the french horn sample from here: http://www.pcabx.com/product/mds-jb920/index.htm (pcabx page), as pointed by Guruboolez

Here (quite noisy environment), I don't hear anything wrong with std4. Is the problem obvious? What should I listen for?

Preecho handling worse with std4
Hihat.wav
Abx Original v std4 11/11
Abx st3 v std4 12/12

I'll check some music this morning but its noisy around here.

French.wav
Original v std4 12/12
std3 v std4 8/8
std3 v original 8/12

I think If I listened closer I could abx std3. With std4 there is a crackling not present with standard 3 its a very odd sound...
You walk away.wav
Std4 v original 8/8
Very chaotic sound the cymbals and bells don't sound right.... Just a very chaotic sound don't know how to describe it.

If you have a sample where std4 is really bad, could you provide the time indication, in order for me to know when is the artifact appearing?

Gabriel, from 4 seconds to the end of the clip I have in an earlier post (You Walk Away From Me) is all warbly with std4. Sounds like cassette tape wow/flutter. It's not subtle, very easy to hear. At least that's the results I'm getting from your compile. std3 doesn't have that problem.

Hmm, I went and encoded my whole folder containing the typical problem samples and none of them have the obvious artifacting that the clip I have produces. But none of them have guitars with heavy distortion, like the Filter sample. One sample you probably have that has pretty clear artifacts is "bassdrum.wav", the cymbals sound distorted with slight but noticeable flanging.

Another sample against standard4, http://www.geocities.com/nyaochi2000/sampl...a-openhihat.zip (Hayabusa by Spitz). Sorry for zip format due to the restriction of geocities.com. And please copy and paste the url in your browser to download it. It contains original(flac), standard1 and standard4 samples.

Listen carefully to two open-hihat beats just after 1 sec. Standard4 corrupts them like old wma(?) although standard1 and other presets work fine.

Thank you...

With those samples, I'm starting to understand why there is a problem...

A bit late : I've uploaded the Miles Davis' sample I talked about in this thread :
http://membres.lycos.fr/guruboolez/AUDIO/s...Dance_short.ape

Standard4 is more than a codec : it's an artist, who like vibrato

Why, if you don't mind me asking?

@guruboolez What kind of equipment do you have?

I listen with headphone only (Philips HP910 & BeyerDynamics DT-531), connected on soundcard (DMX6Fire) for the HP910, and on a small Onkyo amp for the Beyer.

also wondering, what is the config for cbr128?? could it be based on std3 or dibroms in the next build? or is it already?
i found that "Train - Drops Of Jupiter" has major problems on the high-hats throughout, especially on the quiet parts like the intro, at cbr128 using the dll.

/my2pents

@jesseg Drop me a line in my pm or my email box. I have a few lines you could try that may improve cbr128 preformance. May work well for streaming.

@guruboolez How is the S/N ratio of the soundcard? I have a onkyo amp that I do all my dances with and in addition to the backup cd player (I mostly do turntables) I have a computer with a playlist that I will run if I have for any reason to go out. I'm disapointed with the preformance of the audigy I recently purchased (3 days ago) and am taking it back.

Good enough, for my deafening fan...

Official data :

AD signal-to-noise ratio 100 dB
DA signal-to-noise ratio 104 dB

source

Pretty impressive. I'll try it out. It hardly has to be perfect as the echoing off the walls will disguise quite a bit of noise

The Audigy is that bad? I knew it resampled, but I didn't think that beyond that it was that terrible. I was interested in getting an Audigy2, because I know that I can use FB2K's resampler to make my music sound as good as normal, and it has loads of options.

Heh. You would be surprised what is acceptable for small speakers/low amplification. Becomes poor sound using high amplification. Its just I get quite a bit of hiss. Other people have suggested it may be a ground or a computer fan issue. In which case I will use a mp3 player with playlist support. Read:Ipod, Zen

I've found another standard4 problem sample. I think it's the same problem as in guruboolez' Miles Davis sample. Easy to notice in standard4 (ABXed 8/8 in < 1 Minute), with all other settigs not ABXable so far.
It's a salsa song (Original De Manzanillo - Via Libre Que Vienne La Original). The "vibrato" is added to a (synthesized) violin.
If someone wants this sample, PM me - I'll send it as an eamil, as I have no homepage/webspace.

I'm doing some quick testing and will edit the results into this post. I will focus on obvious flaws and won't bother with minimal differences which may be abxed with more carefull listening (or not) since right now we are trying to rule out what clearly doesn't work. On request I can send you the samples, or upload them someplace. I also won't bother to describe the artifacts except if there is something special, since most of the time I would probably be saying "warble".

Sample: Gouryella - Tenshi (original mix) [melodic trance]
Std 0 : can't abx
Std 1 : can't abx
Std 3 : 5/5 easy
Std 4 : 5/5 easy, worse than 3

Sample: Dougie MacLean - Auld Lang Syne [accoustic guitar + male voice]
Std 0 : can't abx
Std 1 : 5/5 pretty hard (guitar strokes give it away)
Std 3 : 5/5 same artifact as in Std 1. Slightly easier to abx (perhaps due to "training" from Std 1)
Std 4 : 5/5 same artifact, even easier to notice. I'm sure this is worse than Std 3.

Sample: Electrasy - Foot Soldiers [Rock with agressive highs]
Std 0 : can't abx
Std 1 : 5/5 highs get smeared
Std 3 : 7/8 highs get smeared. Harder to abx than Std 1.
Std 4 : 5/5 flanging highs very obvious. Worst of the four.

That's all for now