Pre-Test thread, for the 64kbps test Options

[tigre]

Hmm.. for plain AAC (not HE AAC)  I think that Ahead's  64 kbps CBR mode is better than 'Tape' preset.  Also, encoder in the latest Nero6 is better.

Thanks for the info. If I get this right, this means the Nero Mix demo I've downloaded ~1 week ago doesn't contain the latest Ahead AAC encoder - correct? So ... I'll give Nero 6 a try. (Hopefully I get it uninstalled afterwards without any trouble.)

EDIT: I've just downloaded the latest Nero6 and had a look at the extracted files after starting installation: aacenc32.dll is the same version (2.5.5.1) that I had already in my Ahead/Shared folder (HE AAC encoding also works with NeroMix). Is there a newer one I've missed? - I guess not, so I'll start encoding 64kbps CBR ...

Some news: NeroMediaPlayer available for download since 08/29/2003 has a new aacenc32.dll included (v. 2.5.5.2). Besides this problem in a quick test it sounded better than v. 2.5.5.1 (haven't compared to quicktime so far). Hopefully I'll be able to repeat the AAC part of the test this weekend ... The improvement, especially with the worst sounding piano sample was quite big, so maybe we'll get a photo-finish. B)

[tigre]

Test with newest Ahead aacenc32.dll v. 2.5.5.2
same settings as before.

Results:

CODE

---------------------------------------
1R File: .\Samples\003 ChopinPolonaiseDMoll_Nero_vbr_new.wav
1R Rating: 3.0
1R Comment: lower cutoff, ringing and warbeling in some places (e.g. 14-18), bumping echos
---------------------------------------
2R File: .\Samples\003 ChopinPolonaiseDMoll_Nero_cbr_new.wav
2R Rating: 3.5
2R Comment: high tones sound a bit like another instrument, problem with transients
---------------------------------------
3R File: .\Samples\003 ChopinPolonaiseDMoll_qt_44.wav
3R Rating: 3.7
3R Comment: small lack of brightness compared to 2, but less other problems.
---------------------------------------

---------------------------------------
1L File: .\Samples\005 Enola_Gay_qt_44.wav
1L Rating: 3.5
1L Comment: brightness, smeared transients, a little bit of flanging, pre-echo
---------------------------------------
2R File: .\Samples\005 Enola_Gay_Nero_vbr_new.wav
2R Rating: 3.0
2R Comment: less bright than 1, less flanging, a little bit less pre-echo at some places
---------------------------------------
3L File: .\Samples\005 Enola_Gay_Nero_cbr_new.wav
3L Rating: 2.8
3L Comment: same as 3, brighter but serious stereo problem at 1 sec., warbeling
---------------------------------------

---------------------------------------
1L File: .\Samples\006 experiencia_Nero_cbr_new.wav
1L Rating: 3.6
1L Comment: percussion smeared, trumpets suffer from cutoff
---------------------------------------
2L File: .\Samples\006 experiencia_Nero_vbr_new.wav
2L Rating: 2.5
2L Comment: similar to 1; cutoff more noticable, ringing
---------------------------------------
3R File: .\Samples\006 experiencia_qt_44.wav
3R Rating: 3.5
3R Comment: same as 1, trumpets a little bit worse
---------------------------------------

---------------------------------------
1L File: .\Samples\007 gone_Nero_vbr_new.wav
1L Rating: 2.0
1L Comment: 1st part: 2.0 cutoff, warbeling, pre-echo
2nd part: 2.0 warbeling, ringing, flanging, watery
---------------------------------------
2L File: .\Samples\007 gone_Nero_cbr_new.wav
2L Rating: 3.3
2L Comment: 1st part: 3.5 high pitched noise partitially removed + bumping, slightly flanging in some places
2nd part: 3.0 ringing, flanging, cymbals smeared
---------------------------------------
3L File: .\Samples\007 gone_qt_44.wav
3L Rating: 3.5
3L Comment: 1st part: 4 high pitched noise partitially removed
2nd part: same as 2
---------------------------------------

---------------------------------------
1L File: .\Samples\009 mybloodrusts.sample20sec_Nero_vbr_new.wav
1L Rating: 2.3
1L Comment: 1st part: 2.5 stereo problem
2nd part: 2: loud ringing
---------------------------------------
2R File: .\Samples\009 mybloodrusts.sample20sec_qt_44.wav
2R Rating: 2.5
2R Comment: 1st part: 3.5 small stereo problem (chirping)
2nd part: 1.5 constant chirping, ringing
---------------------------------------
3R File: .\Samples\009 mybloodrusts.sample20sec_Nero_cbr_new.wav
3R Rating: 2.3
3R Comment: 1st part: 1.5 annoying stereo problem
2nd part: 3 guitar+cymbals smeared but no annoying ringing
---------------------------------------

---------------------------------------
1R File: .\Samples\011 Scars_Nero_cbr_new.wav
1R Rating: 3.8
1R Comment: percussion smeared
---------------------------------------
2R File: .\Samples\011 Scars_qt_44.wav
2R Rating: 3.8
2R Comment: same as 1
---------------------------------------
3R File: .\Samples\011 Scars_Nero_vbr_new.wav
3R Rating: 2.5
3R Comment: percussion smeared, lowpass, ringing, warbeling
---------------------------------------

---------------------------------------
1L File: .\Samples\012 Waiting_Nero_vbr_new.wav
1L Rating: 2.0
1L Comment: 1st part: stereo separation OK but annoying ringing
2nd part: smeared, flanging, ringing cymbals
---------------------------------------
2R File: .\Samples\012 Waiting_qt_44.wav
2R Rating: 3.0
2R Comment: 1st part: 3 same as 1 but better
2nd part: similar as 1 but less lowpass
---------------------------------------
3L File: .\Samples\012 Waiting_Nero_cbr_new.wav
3L Rating: 3.2
3L Comment: 1st part: 3.2 same as 2 but slightly better
2nd part: as 2
---------------------------------------

Conclusion:
Ahead aacenc CBR and Quicktime seem to be equal if there weren't the stereo issues. As there are at least two severe problem cases for Ahead aacenc among the samples chosen, Quicktime is the only option.

[rjamorim]

Great! Thanks for the information, and for taking the time to perform these tests

This post has been edited by rjamorim: Aug 31 2003, 00:04

[bond]

So, here's an updated list of codecs to be featured:

-Nero HE AAC
-Ogg Vorbis
-MP3pro
-WMAv9 std
-Apple AAC LC
-Higher anchor: Lame --alt-preset 128
-FhG MP3 64kbps w/o IS

Real is on hold due to reasons out of my league. I'll know if it'll be featured or not by the weekend.

imho i would test:
- Nero HE AAC
- Ogg Vorbis
- MP3pro
- WMAv9 std
- RealCook (hey, its the direct competitor to wma and used very often!)

- anchor: Lame --alt-preset 128

-> 6 codecs, nice amount for testing, interesting results also for "the masses"...

[askoff]

imho i would test:
- Nero HE AAC
- Ogg Vorbis
- MP3pro
- WMAv9 std
- RealCook (hey, its the direct competitor to wma and used very often!)

- anchor: Lame --alt-preset 128

-> 6 codecs, nice amount for testing, interesting results also for "the masses"...

That list looks bit better for me too. I think i would also add Aplle AAC LC with them.

[rjamorim]

I was thinking of creating a poll offering people several options for a codec suite.

But then, I would have to postpone the test for at least a week.

Decisions, decisions... >_<

[askoff]

I was thinking of creating a poll offering people several options for a codec suite.

But then, I would have to postpone the test for at least a week.

The poll idea would be great. More happy people==less whining.
I hope we are no hurry on this test. I remember many stupid finnish proverb about doing something in hurry...

[bond]

just an info:

according to JohnV vorbis 1.0.1 "is expected in few days actually"

[JohnV]

just an info:

according to JohnV vorbis 1.0.1 "is expected in few days actually"

Well.. according to #vorbis irc-channel topic actually..

[rjamorim]

OK.. So, the million dollar question is: Will it make any difference at my test? Is low bitrate tweaking planned for this release?

[bond]

no quality tweaks, just bug fixes as far as i know, but i think you can be sure that people will start whining if you dont use it

[JohnV]

OK.. So, the million dollar question is: Will it make any difference at my test? Is low bitrate tweaking planned for this release?

[05:29] <xiphmont> 1.0.1
[05:29] <xiphmont> all the CVS fixes that have been building up, as well as bugs reported in other forums.
[05:29] <xiphmont> ...as it turns out there will be a few tuning fixes there too.

So.. few tuning fixes.. Hard to say if those would make any real difference though.

[rjamorim]

but i think you can be sure that people will start whining if you dont use it

Right... :B

OK, test start is postponed for one week. New expected dates are Sept 10th to 21st.

[bond]

more info about 1.0.1 here

[dev0]

Improved handling of quiet signals in low bitrate modes

This could be quite significant for the test, so waiting is the right decision.

dev0

[phong]

This is a relavent thread to the vorbis fix.

[ff123]

ABC/HR 0.9b doesn't save the results when I try to setup a new test.
I get a blank results file with the title of the new test!
I had to do the test twice (except Waiting) because of the bug.

Bug noted and added to the short list. I will try to release a new version soon which obscures results (not for this test though).

[998 spams/virii in 18 days of vacation, including about a dozen real email].

ff123

[rjamorim]

Welcome back, master!

[rjamorim]

OK, the test will start this Wednesday, no matter if Vorbis 1.0.1 is released or not.

So, I'll lay out the general headlines of what will be tested:

Samples: The ones mentioned at the first post in this thread:
http://www.hydrogenaudio.org/show.php/showtopic/12358

The samples suite is now frozen and won't change.

The encoders that are planned to be featured are:

HE AAC from Nero 6.0.0.15, Streaming :: Medium, High Quality
Ogg Vorbis 1.0.1 or post-1.0 CVS, -q 0
MP3pro from Adobe Audition 1.0, quality 40, Current codec, allow M/S and IS, allow narrowing, no CRC
Real Audio Gecko/Cook 64kbps from Real Producer 9.0.1
LC AAC from QuickTime 6.3, Best Quality
WMA v9 VBR quality 50
High anchor: Lame 3.90.3 --alt-preset 128
Low anchor: MP3 from Adobe Audition 1.0 (FhG), 64kbps CBR, Current codec, allow M/S, no I/S, allow narrowing.

About resampling: I won't resample anything prior to encoding, but if the encoder resamples by default on that specific setting, I won't force it to use another sample rate. Therefore, for instance, FhG MP3 will end up with 22050 kHz, since that's the default in Audition.


Now, something important that MIGHT happen: I'm talking to a codec developer, and tomorrow I'll receive his reply if they want me to test their codec or not. IMO, that codec is very worth testing, as from what it seems, it might well be the winner, or be among the winners. I can't speak more about it now because I still have no answer from him.

So, if this developer gives me the thumbs up, what codec you guys think should be replaced by it? I'm leaning towards LC AAC or MP3, but it's up to you. Any opinions?

Thanks.

Regards;

Roberto.

[askoff]

I dont think that MP3 at 64kbps is needed because it surely will be the looser of all these codeks.

[elmar3rd]

IMO, that codec is very worth testing, as from what it seems, it might well be the winner, or be among the winners. I can't speak more about it now because I still have no answer from him.

So, if this developer gives me the thumbs up, what codec you guys think should be replaced by it? I'm leaning towards LC AAC or MP3, but it's up to you. Any opinions?

Is it related to the rumors over mp3 with mpc-psymodel? Whatever, i would replace one of the anchors, as the comparison of the codecs is much more important than anchors, imho.

[Dibrom]

Is it related to the rumors over mp3 with mpc-psymodel?

I very seriously doubt it. I also doubt that this codec (the mpc/mp3 thing) will ever see the light of day in the capacity that it was originally rumored it might.

[rjamorim]

Is it related to the rumors over mp3 with mpc-psymodel?

Nonono. That project is most probably dead.

And, even using MPC psymodel, there is no way MP3 could be "among the winners", like I said.

[tigre]

Although I have invested some time into it, I'd recommend to replace mp3@64kbps. From the codecs I've tested @64kbps recently it has the lowest lowpass which will be the most obvious difference to notice with the majority of samples. The other codecs tested will have most problems with "real" encoding artifacts (ringing, pre-echo, changed background noise, etc), so this lower anchor probably wouldn't be worth that much anyway.

[brett]

mp3 = part of the past.

lc aac = part of the future.

(and i thought you agreed that qt's performance in the last two tests (a win and a strong second with the vbr millstone) logically demanded its inclusion in this test. no?)