64kbps listening test, OPEN Options

[nwn]

null

This post has been edited by nwn: Sep 14 2003, 11:17

[ff123]

YES Exactly ...
i also try to launch soft and whithout load config ... select direct file/setup test
browse for the orignal file and the first wave without changing offset

exact blue icon activate ...

no change on time ...

adjust offset :
Offset_Orig =  14
Offset1  =  0

and no more change ...

I test from scratch with the same file for orig and first ....
offset 0 .. then 14
no more result....

ok on the batch process ...
wav04 is Lame
wav07 is faad

for Lame i've 8 lame.Exe and 5 different versions of lame on some different dir .... and the binone in more .. with a prefetch lame ... must fix that ...
the lame one in bin is v3.903mmx
is there a path problem ..

Faad two file same size 13K don't know release no seen under dos ..

nwn,

I'm having trouble understanding exactly what is and isn't working in your last message. Please clarify. You may have to supply pictures of either the setup window or a copy of whatever config file it is you're using.

ff123

This post has been edited by ff123: Sep 14 2003, 03:29

[rjamorim]

nwn;

Comme une exception, je crois que tu peux écrire en français et ignorer, seulement cette fois, la règle de "seulement anglais".

Après, moi ou Pio2001 peut traduire ton post par l'anglais.

---------
For those that want to understand WTF I wrote:
--
As an exception, I think you can write in French and ignore, only this time, the rule of "only english".

After, me or Pio2001 can translate your post to English.
---------

Roberto.


OBS: Please noone come whining rubbing rule 10 on my nose. This is a special case. I hope you understand.

This post has been edited by rjamorim: Sep 14 2003, 06:45

[wkwai]

I noticed that people posted audio clips which are too loud.. You should not use loud clips for subjective listening tests.
This has something to do with the masking effects. For loud tones, the spreading functions for the masker is not as steep as those of quieter tones.
You would probably noticed that you can hear more "hidden tones" when the volume of your player is lower than when it is higher.
This effect is modelled in the ISO Psychoacoustic Model I.

[nwn]

okay ...

I did exactly what fff123 said ..

first i decide to start from scratch again ..
re-download from http://audio.samharris.us/test/abc-hr_bin.zip
- decompress in C:\Downloaded\abc-hr_bin
- run abc.Exe
- i didn' load any config
- go to "file/setup test"
- browse "orig wave" and first "wave" to the same file

"BigYellow.wav"
bitrate = 1411
samplerate = 44100
channels = 2
codec = PCM
bitspersample = 16
----------
1080185 samples @ 44100Hz

- no change offset
- click ok ...
- "blue" arrow appears .. ok
- bar displayed in gray .. but OO (start) ---- OO (stop)sec
SO what button I click ...NO SOUND ....

- go to setup test
- change offset to "Offset_Orig = 14"
- press Ok
- same result ........

I decided to test with another pack ..
downloaded "64kbsGroupTest" with NewKid
Same effects ...

same display than i post


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

when i looked on the file infos ...
i saw than bitrate and bitspersample different for
wav04 is Lame encoded
wav07 is faad encoded

so .. i looked for lame.exe and faad.Exe
- lame.exe in \bin dir is v3.903mmx (but i have some others lame.exe in other dir)
- faad.Exe in \bin dir is 134k size

________________________________________________

okay ..
- i didn't move any file or dir ...
- i loaded all the packs and run the batch for each
- same results for each pack
- no other error message ...

is this post more explicit and in a better english?

what do you want
- i do
- screen capture
- i verify

[tigre]

I noticed that people posted audio clips which are too loud.. You should not use loud clips for subjective listening tests.

There are only loud clips in > 90% of mainstream music of the last 10 years (mastered at high volume). A test that uses clips that have nothing to do with real life wouldn't be worth much, would it?

This has something to do with the masking effects. For loud tones, the spreading functions for the masker is not as steep as those of quieter tones.
You would probably noticed that you can hear more "hidden tones" when the volume of your player is lower than when it is higher.

I don't understand where's the point. Audio players that deserve this name have features to change volume (= e.g. volume slider). Noone forces you to do the test at a certain volume.

As you probably have noticed, the test has already been started an there was a pre-test thread open for discussion for some weeks (including the clips). It might be a good idea to mention your concerns earlier next time.

[pseudoacoustic]

I'm unable to replicate this problem with the "gone" package, by the way.

ff123

Me either. They've all worked so far.

[nwn]

me too unable to reproduce on others computers ....

in fact my reputation must be saved ...

i decide a great .. decision .. to test on one of my five other computers in home ..

on two others PC .. after install ... ALL is working fine ....

so this is my stupid pc station that have some issue ..
multi-layered install/desinstall for more than two years ... on xp ... standard issue ..

in fact no icon sound wants to be displayed on taskbar and my volume key on Ms MM Keyboard are producing noting ..

it's some sample of my dirt configuration ..

no problem . as my hard disk is going to have TEC issue by november .. (smart monitoring) i will reinstall a new one .. but no ibm 60/75 series of course ..

sorry ... stop search to resolve this issue
this is probably non chance

This post has been edited by nwn: Sep 14 2003, 15:42

[2Bdecided]

Glad you found the problem.


Everyone: go on, try this test - it's fun!

Cheers,
David.

[Atlantis]

Everyone: go on, try this test - it's fun!

I completely agree.
That was my first listening test at this bitrate, and while some samples are extremely easy to abx / test, a pair of them are really tricky (what a headache! )

I can't wait for the final results (and if they are like mine... big surprises )!

This post has been edited by Atlantis: Sep 15 2003, 11:34

[joey_m]

Everyone: go on, try this test - it's fun!

... or frustrating! Maybe because of my sub-par gear or ears (probably both!), but some samples are giving me quite a hard time!


Cheers, Joey.

[Atlantis]

Everyone: go on, try this test - it's fun!

... or frustrating! Maybe because of my sub-par gear or ears (probably both!), but some samples are giving me quite a hard time!


Cheers, Joey.

Nah

That's the fun!

[Gambit]

Roberto: While doing the test I've noticed that the files have different sample rates.
Of course, that's because each encoder chooses the appropriate sample rate at this bitrate.
But wouldn't it be better if all the samples would be resampled to 48kHz after decoding? To avoid resampling issues etc...

This post has been edited by Gambit: Sep 16 2003, 15:56

[AstralStorm]

I think we're testing real-life situations, SB Live/AC97 playback too.

/EDIT\
Yeah, I had some hard samples too, mainly because these are no problem samples,
LAME AP128 (I think these were compressed with it) doesn't damage them too badly
and I did the test at 0300 UTC. (it's night here then)

Now ABXed all my 5.0 samples and it wasn't very hard.
\EDIT/

This post has been edited by AstralStorm: Sep 16 2003, 17:00

[2Bdecided]

I think we're testing real-life situations, SB Live/AC97 playback too.

Yes, but what about cards which sound very different at different sample rates, or glitch when changing sample rates? It makes ABX a bit pointless! And could make the test quite unfair.


Luckily the Audiophile 2496 changes between 22.05, 32 and 44.1 seamlessly, whereas it mutes and fades in when switching to 96k (and back).

Cheers,
David.

[AstralStorm]

Yes, but what about cards which sound very different at different sample rates, or glitch when changing sample rates? It makes ABX a bit pointless! And could make the test quite unfair.

Only issue is with 32kHz, as some people might not detect lowpass. (16kHz)

But what about cards supporting only 44,1kHz, like some AC97's?

Anyway, you shouldn't need to ABX these samples anyway - they're damn easy (maybe except few samples nr 5 and these are 44,1kHz anyway)

This post has been edited by AstralStorm: Sep 16 2003, 17:07

[rjamorim]

Roberto: While doing the test I've noticed that the files have different sample rates.
Of course, that's because each encoder chooses the appropriate sample rate at this bitrate.
But wouldn't it be better if all the samples would be resampled to 48kHz after decoding? To avoid resampling issues etc...

Well, I already discussed that with Darryl, because tigre also brought this issue some days ago. There are some points we came up with:

- In a sound card that resamples everything to 48, all samples will be treated equal - I.E, all will be upsampled since there are no 48kHz samples

- Test would need to be restarted! I can't mix together the results of a non-resampled test with the results of a resampled one. People would hate me for that.

- Someone might complain that "no, my favorite codec WAS performing well, but then that stupid Roberto added resampling, and everybody knows resampling reduces the quality". Specially if this test is announced at /. :B

- On some very old cards (like the SB16 I owned for years and now is owned by my mother), if you feed them 48kHz signal the quality gets terrible unless you change the resampling quality setting at the control panel (the default is lowest quality). I remember all the hissing when I tried playing my first DVD here. >_<
So, it's also hard to decide upon a standard sampling rate for all samples.

So, if people think this is a major issue, that's OK. I'll restart the test. But there's no way I can change the settings now, halfway through it.

Thanks for bringing this issue.

Regards;

Roberto.

[2Bdecided]

We don't discuss whether they're easy or not until the test is over.

But look further up this thread at some of the comments...


I had a sound card that _didn't_ resample - it just inserted duplicate or zero samples. It sounded horrible. Should I have graded 22kHz sampled material using this card? Actually, you should have heard 11kHz sampled material!

But that was a pro, the DAL CardD - one of the best at the time.

Cheers,
David.

[2Bdecided]

No no no - don't re-start the test!

Just remember it for next time ( ), and mention it in the results. It probably doesn't matter that much, if at all - just being rigorous.


btw, it would be natural to resample everything back to the original sample rate, e.g. 44.1 in this case. Otherwise you're comparing sound cards and windows resampling and codecs, rather than just codecs!

Cheers,
David.

[Gambit]

I also don't think that there is a need to re-start the test. I just brought it up as an idea to be considered for future tests.

[rjamorim]

I also don't think that there is a need to re-start the test. I just brought it up as an idea to be considered for future tests.

OK, thanks.

I'll make sure I'll mention it at the results page as well.

[schnofler]

Hi,
this is my first post to this forum, and I don't know whether that is the right topic to post this to, but I guess you might be interested in this so I just post it here.

So, let's get to the point. I programmed a Java version of ff123's ABC/HR application so all those Linux users out there can participate in listening tests in the future (well, actually I just had nothing better to do, but that doesn't sound so philanthropic )
It's still a very early alpha (I started development five days ago), but it's already pretty functional. I designed it to be as compatible as possible to ff123's Windows application. It reads the same config-files and writes the same result files, and overall my Java version is pretty much a straight rip-off of the original. It also has some additional features like optional disabling of the fast-switching feature, looping and saving sessions in progress.

Now two issues arise: First, I'd like to ask ff123, whether you mind me publishing such a blatant copy of your work, and, if you don't object, what exactly I'd have to do so I don't violate this Lesser GPL under which you distribute your application (I don't have a lot of experience on these things, and I didn't feel like working through those ten pages of legal mumbo-jumbo in the license.txt).
And secondly, I don't have a website or webspace myself, so if some of you are interested in my version, it would be nice if one of you could make it available to the public (the program is only about 100KB in size).

Thanks in advance, and also many thanks to rjamorim for organizing this great listening test.

[rjamorim]

Hi,
this is my first post to this forum, and I don't know whether that is the right topic to post this to, but I guess you might be interested in this so I just post it here.

So, let's get to the point. I programmed a Java version of ff123's ABC/HR application so all those Linux users out there can participate in listening tests in the future (well, actually I just had nothing better to do, but that doesn't sound so philanthropic  )
It's still a very early alpha (I started development five days ago), but it's already pretty functional. I designed it to be as compatible as possible to ff123's Windows application. It reads the same config-files and writes the same result files, and overall my Java version is pretty much a straight rip-off of the original. It also has some additional features like optional disabling of the fast-switching feature, looping and saving sessions in progress.

Now two issues arise: First, I'd like to ask ff123, whether you mind me publishing such a blatant copy of your work, and, if you don't object, what exactly I'd have to do so I don't violate this Lesser GPL under which you distribute your application (I don't have a lot of experience on these things, and I didn't feel like working through those ten pages of legal mumbo-jumbo in the license.txt).
And secondly, I don't have a website or webspace myself, so if some of you are interested in my version, it would be nice if one of you could make it available to the public (the program is only about 100KB in size).

Thanks in advance, and also many thanks to rjamorim for organizing this great listening test.

Woot!

Your tool is a Godsend

About licensing: trying to make a loong license short, the LGPL states that if you change LGPLd code, you must make the changes available somehow. That's all.

It doesn't force you to open the sources of all the work related to it, like the GPL.

And I would be very glad to host it at RareWares (check the bottom of the post). ff123 probably would also be happy to host in his ABC/HR page.

Can you mail it to me?


Thanks a lot for your effort

I'll test it ASAP and, if there are no issues/bugs, I'll use it as an option for Linux/Mac/whatever users.

Best regards;

Roberto.

One question: Out of curiosity, what runtime engine version is needed to run it?

[schnofler]

OK, I mailed the program to you. I doubt there won't be any bugs, though, so if you find some, I'll be glad to eliminate them
The JDK I used is 1.4.2, but generally it should run on 1.4, maybe even 1.3 (Sun did quite a number of bugfixes to the Java Sound API since then, however).

[ff123]

So, let's get to the point. I programmed a Java version of ff123's ABC/HR application so all those Linux users out there can participate in listening tests in the future (well, actually I just had nothing better to do, but that doesn't sound so philanthropic  )
It's still a very early alpha (I started development five days ago), but it's already pretty functional. I designed it to be as compatible as possible to ff123's Windows application. It reads the same config-files and writes the same result files, and overall my Java version is pretty much a straight rip-off of the original. It also has some additional features like optional disabling of the fast-switching feature, looping and saving sessions in progress.

That's great!

If Roberto hosts the program on his site, I'll just create a link on my page to point to it.

I'm in the process of writing code to "obscure" the config and results files, so when I get that worked out, I'll let you know. But since the code will be freely available, it shouldn't be a problem to duplicate.

Looks like I'll also have to add in the looping capability and disabling fast switching as well, so that the applications stay as similar as possible.

Could you email me a copy as well? (my email is in the "about" box of abc/hr)

ff123