Hello.
I'd like to announce the launch of my 32kbps dial-up bitrate listening test

The formats featured are Nero Digital Audio (HE-AAC+PS), Ogg Vorbis, WMA9 Std., MP3pro, Real Audio and QDesign Music Codec.
Lame MP3 is being used as low anchor, and a lowpass at 7kHz is being used as high anchor.

For further information on how to participate, visit the test announcement page:
http://www.rjamorim.com/test/32kbps/presentation.html

The test is scheduled to end at July 11th.

Thank-you very much.
Best regards;
Roberto Amorim.

the faad version for OS X on rarewares appears to decode the HE-AAC samples to mono files. Well I only tested the new york one but it was the case there.

Isn't mono the best alternative for such low bitrates?

Maybe, but the test was HE-AAC+PS, and PS is useless on a mono file by definition, so I doubt the samples were mono.

Please make sure the files are decoded with recent FAAD2 on all platforms as there have been updates/bugfixes to HE-AAC decoding.

No recent version of FAAD2 should give mono output on *any* file BTW.

Hrm... I guess RareWares needs an updated compile of Faad.

Can any Mac user contribute one?

The torrents seem to be flowing well Are you going to bother sending the word out to the unwashed masses at Slashdot et al? 50gbs of bandwith leeched for ~0 participants seemed futile last time round

Might be worth it. But I don't think they would be interested in a 32kbps test. And, of course, we would se hundreds of posts there claiming such low bitrates are useless, yadda yadda.

Atill, it could bring some participants. I would be grateful if someone could post the news there.

(off topic ?)
As a newcomer to listening tests (my first ! ), I fear my comments will miss some technical vocabulary - what about a library of audio artifacts with 'technical' terms associated ?
Anyway such a test seems interesting and promising. Thx !

For that matter, is full 44 khz stereo necessary at 32 kbps? If I were streaming 32 kbps, in the interest of quality I would probably downsample to 22khz mono, or even 11 khz mono.

Only Nero AAC uses full 44.1kHz

All other codecs had the samples downsampled for encoding, and later upsampled for comparision on ABC/HR

The reason of upsampling is working around crappy sound card resamplers, that would bias the test.

I did 4 samples tonight. They are rather easy to do because you don't have to strain to differentiate between the sample and the reference. The results should be very interesting.

Downloaded everything and have it all unzipped an ready to go, but I'm holding off on it right now because I'm sleepy I'll be pretty worried though if I can't tell 32kbps files from the source lol. Seriously I am looking forward to the testing.

I shouldn't be surprised, but harashin again finished the entire sample set first. This time in less than 12 hours after the test start.

He sure needs real life then.

I haven't been keeping up with this test but can someone tell me why HE-AAC with PS is being used, MP3 Pro but only WMA standard is being used not Pro? doesn't seem fair to me

Because the Pro-codec doesn´t work with low bitrates, AFAIK only >= 128 kbps.

For the Unix enthusiasts, create theses scripts in the bin directory of the test (!! case sensitive !!):

Samplexx.sh


Sample.sh


And run with

Please make sure that for Unix tests you use latest FAAD2 (www.audiocoding.com) - otherwise some strange things could happen

From my gentoo, the ebuild is "media-libs/faad2".

I prefer if a version build from yesterdays checkin would be used. But this should not give any big problems.

Menno

To all extra-lazy windows users who downloaded the "all samples" package and don't want to execute those sampleXX.bat files one by one:

I made a batch file for converting all samples at once, just download it to your test/bin directory and execute it (of course all sampleXX.zip files must have been extracted to test/SampleXX before).

download link

Regards, fileman.

Finished all 18 samples.

Results submitted.

G'night everybody.

likewise and good day/evening.

I added your scripts to abc-hr_bin.zip. Hope you don't mind.

Thank-you very much.

Regards;

Roberto.

Not at all.

Hello.

I would like you to know that, due to an "emergency" (I must go to a $^%!@ cousin's wedding next weekend. My parents and my brothers won't be able to go, so I must go to represent our side of the family), the results won't be published with the usual speed I'm so proud of.

With a little luck, I'll be able to release them on monday afternoon. But don't hold your breath. Anyway, I'll find a way.

Sorry for the incovenience.

Regards;

Roberto.

Don't worry about that - everybody knows that you're doing a great job with all these tests! Enjoy the wedding, eat some nice food

http://slashdot.org/article.pl?sid=04/07/0...tid=141&tid=188

WEEEEEEEEEEEEEEEEE


@fileman: thanks

I'll take the blame for that. I am CaptainCheese! You asked, no-one objected...

I winder if anyone from /. will actually help out...

Hi,

Not in VBR Quality mode. 10%, 25% and 50% quality are low bit rate profiles for WMA 9 Professional and the average bit rate can be reduced to as low as 45 Kbps.

But we still cannot use WMA9 Pro because it is Quality VBR, the bitrate is highly fluctuating and doesn't stay at 32 Kbps.

Cheers,
McoreD

Yes, that's OK. There are probably a handful of /. users that actually participate.

All is fine, as long as I ignore most of the comments...

Thanks for announcing it there.

I'm trying to run the test on OSX. When running Sample.sh, I get this segfault

Samplexx.sh: line 16: 15841 Segmentation fault ssrc --rate 44100 --twopass $1_1a.wav $1_1.wav
Shibatch sampling rate converter version 1.29

Samplexx.sh: line 17: 15844 Segmentation fault ssrc --rate 44100 --twopass $1_2a.wav $1_2.wav

I compiled ssrc from source, using no special options. Any help would be appreciated!

--Andrew

Hi, I'm wondering what Real Audio Codec is used in the test? Because they are using different technology. CELP/Subband/Transform

If the codec from Real is CELP based it should win the speaking samples.

Anyway the codecs are in theise groups, maybe some results can be connected to them later:

CELP:
Real (if it is Sipro/Cooker)

Subband:
Lame
Mp3Pro

Transform:
AAC (DCT)
Vorbis (DCT)
QDesign (FFT)
WMA9 (DCT)

Tip to x86 unix users. Its much easier to use wine than download and compile a load of software. Just type:

wineconsole Sample01.bat

and so on for each sample. It uses the windows EXEs included with abchr and works pretty well.

Don't do what I did and spend loads of time getting the programs to work with the shell scripts only to discover my faad was not recent enough after taking the test!

MP3 is a transform format too.

Anyway, the speech samples produce good results with one codec, and very bad with another one...

Any news on faad for macosx? The one on rarewares is dated 2003 it says, so I guess that's too old for this test...

Right. Noone contributed new FAAD though :/

So it's officially over now ? I can't wait for the results

Someone can give the GMT cloture hour?
i've made the test for on sample (debussy 03) but i've sent my result one hour ago and i don't know if it wan't too late...

Hmm ?
I may be wrong, but mpeg layer 3 is a tranform codec.
As for subband codec examples: mpeg layer 2 and mpc...

EDIT: Half of test results I sent were w/o a name, half with it. Sorry for that, Roberto, if it will raise any problems...

So ? No news ?
At least that means Roberto is hard working on it

Wish me luck. It'll still take some time, but hopefully I'll get access to a laptop and dial-up later today at my aunt's home.

That's OK. I didn't even grab all results from my Yahoo mail (there are about 40 result e-mails there that need to be fetched, decrypted, renamed, sorted into folders...) so, by all means, results are still being accepted.

Good luck then

Welcome to my daily internet experience

I'm looking forward to the results of this latest test. Good luck getting things together!

Although it might be a bit late with the test already past its end, could anyone explain what JABCHR is completely failing to do anything on my computer?

The biggest of its problems is that no sound plays, kind of unhelpful for an audio app All of the play buttons just do nothing. I don't believe it attempts any playback, from what I can tell the position slider doesn't move at all. I'm not entirely sure as it doesn't display quite correctly, everything has a new found desire to hide from me The app loads completely blank, but resizing the window to completely nothing then expanding again makes everything but the slider appear. As other windows obscure it anything that was hidden becomes blank after it is exposed again.

Everything used to work fine, but I have reformatted and changed some hardware - I am now using the onboard AC97 on my Via motherboard, using their latest Vinyl Stylus drivers. Everything else can play sound quite happily, it's just JABCHR that fails. I also downgraded my graphics card a bit, but it is using the drivers Windows assigned it quite happily for every other task.

Eh?!

Well, which version of Java? Below 1.4.2 there will be problems,
as Java Sound still is in development.

It sais v1.4.2_03 - the same version I was running before.