New Public Listening Test (July-August 2011), Discussion on preparation Options

[IgorC]

Winamp, starting with v5.62 (released today)

Great news.

Plan:
Decide what to test (AAC or Multiformat) and have an idea what codecs - until July 1. It's DONE.
Bitrate verification - 7-10 days - until July 7-10
Sample selection - 7-10 days - until July 15-20
Checking all conditions, preparations, dummy packages etc. Developers can also take a look if everything is alright. - until July 23-24
Start of test - last week of July.

http://listening-test.blogspot.com/2011/06...reparation.html
The period of voting for AAC or multiformat public test is over.
There are more votes in favor of AAC test. The discussion on multiformat test will be continued in next year.

The list AAC encoders and first approximation of the settings for ~96-100 kbps:
1. Nero -q 0.345
2. QuickTime [TVBR and/or CVBR (?)] . The members suggest what settings to use during bitrate verification.
qtaacenc:
TVBR: --tvbr 46 --highest --samplerate keep
CVBR: --cvbr 96 --highest --samplerate keep or iTunes CVBR equivalent : --cvbr 96 --normal --samplerate keep
3. FhG VBR 3 (Winamp 5.62)
4. Coding Technologies (Winamp 5.61). In case if most of the codecs will produce around 100 kbps on average (more likely) then we can use MediaCoder. It can encode with CT at 100 kbps

Low anchor is ffmpeg AAC at 128 kbps. (very inferior quality)

People, suggestions, please.

P.S. Everybody is welcome to #hydrogenaudio on irc.freenode.net

This post has been edited by IgorC: Jul 1 2011, 02:39

[Garf]

Vote for AAC test.

And I also have a question: what software uses/will use new FhG encoder?

Winamp, starting with v5.62 (released today)

Can you make a news submission (specifically the new encoder)? This is interesting enough news.

[IgorC]

It's important to mention that Quicktime had (or has ) a bug at 96 kbps. It appears like lowpass wasn't applied properly and there were some quality drops.
If encoded file is lowpassed at 15.7 kHz then everything is ok.

This post has been edited by IgorC: Jul 1 2011, 19:43

[romor]

I tried nero (1.5.4), FgH (Winamp 5.62), QT (7.6.9) encoders on 2 releases: http://i.imgur.com/Xs8t6.png

That's not much, but knowing what kind of music is behind this releases I might expect that TVBR 46 would be min at average and nero max in larger set of tracks, with recommended parameters

If we consider boxplot of release bitrates as relevant, then QT TVBR seems like right choice over CVBR: http://i.imgur.com/1sVYy.png

Again this are only two unrepresentative releases, (one is noise/microsound and other highly compressed d'n'b), but as I made it out of my curiosity (along with spectrograms etc ) I thought to post it here also

[IgorC]

Thank You, romor.

That's exactly what we should do. Everybody reports his/her resulting bitrates and suggestion about CVBR/TVBR choice. It can be 1 album or whole collection.

I will create Excel table later.

This post has been edited by IgorC: Jul 2 2011, 03:40

[Steve Forte Rio]

My quick test:

Karl Bohm, Wiener Philharmoniker / Mozart: Requiem d-moll, KV.626

Nero 1.5.4.0 -q 0.345 -ignorelength - 96 kbps avg

QAAC v0.47 (QuickTime 7.6.9) -s -V 46 -q 2 - 99 kbps avg (QAAC's resultant audio stream is bit-identical to QTAACENC --tvbr 46 --highest --samplerate keep)

But as for me there are also audible (and visible) diferences in encoding of high frequencies (while the lowpass filter frequencies are nearly equal). I can hear that for QT background noise has a little wider frequency range:

CODE

foo_abx 1.3.4 report
foobar2000 v1.1.7
2011/07/02 11:15:26

File A: D:\Моя музыка\Mozart\Mozart - Requiem [Mathis, Hamari, Ochman, Ridderbusch - Karl Bohm] (1990) - DG\03 Tuba mirum_qaac.m4a
File B: D:\Моя музыка\Mozart\Mozart - Requiem [Mathis, Hamari, Ochman, Ridderbusch - Karl Bohm] (1990) - DG\03. Tuba mirum_nero.m4a

11:15:26 : Test started.
11:15:58 : 01/01  50.0%
11:16:03 : 02/02  25.0%
11:16:10 : 03/03  12.5%
11:16:21 : 04/04  6.3%
11:16:27 : 05/05  3.1%
11:16:39 : 06/06  1.6%
11:16:52 : 07/07  0.8%
11:16:55 : 08/08  0.4%
11:16:59 : 09/09  0.2%
11:17:02 : 09/10  1.1%
11:17:06 : 10/11  0.6%
11:17:13 : 11/12  0.3%
11:17:20 : 12/13  0.2%
11:17:32 : 13/14  0.1%
11:17:36 : Test finished.

----------
Total: 13/14 (0.1%)

Here is the spectrogram of two tracks (the first part is for QT, second - for Nero).

This post has been edited by Steve Forte Rio: Jul 2 2011, 09:26

[IgorC]

@Steve Forte Rio,
Ok, I see QuickTime works fine on your machine (no bug).

Can I ask you to encode a few albums (different musical genres) with Nero q0345, CVBR 96, TVBR 46, Fhg VBR 3 for bitrate verification? Winamp's CT is CBR so no need to verificate it. Tell me if these settings produce similar bitrate (100 kbps) and if some of them need any correction.
Also it's important to check real file size instead of foobar's bitrate information. Sometimes it doesn't inform correctly.


As I'm (one of) conductor of the test it will be too much control for one person (me) so I shouldn't verificate bitrate. I will rely on your and other members results of bitrate verification.

This post has been edited by IgorC: Jul 2 2011, 09:39

[Steve Forte Rio]

Can I ask you to encode a few albums (different musical genres) with Nero, CVBR, TVBR for bitrate verification?

Ok , I'll do it a little later and will report a file-based bitrate (with tags removed) here.

[IgorC]

Thank You.

[hellokeith]

What are the "default" settings for AAC 96kbps on Itunes OSX and Itunes Windows?

[Steve Forte Rio]

Well, here is my results of bitrate verification.

Unfortunately, the variety of genres isn't too big:

Mozart - The Late Symphonies, CD1 (symphonic)
Mozart - Requiem D-moll KV626 (choral)
Bach - French Suites, Vol.1 (solo instrumental)
Led Zeppelin - Led Zeppelin IV, Coda (rock/n/roll)
The Beatles - One (rock/n/roll)
VA - Golden Jazz (jazz)
Bee Gees - The Greatest Hits, CD2 (pop)
Armin van Buuren - Mirage (electronica/trance)

Total duration: 9:38:22 (34702 seconds)

All sources are true lossless

Duration of resultant files for all coders except Winamp FhG is identical to the original. Seems like FhG can't encode gaplessly - I've got +3 seconds totally for it.

Bitrates:

PCM WAV: (6 121 423 616 bytes / 34702 seconds) * 0.008 = 1411.2 kbit/s

[TVBR] QAAC v.0.47 (QT 7.6.9) -V 46 -q 2: (422 549 458 bytes / 34702 seconds) * 0.008 = 97.4 kbit/s

[CVBR] QAAC v.0.47 (QT 7.6.9) -v 96 -q 2: (441 005 435 bytes / 34702 seconds) * 0.008 = 101.7 kbit/s

Nero 1.5.4.0 -q 0.345 -ignorelength: (441 633 426 bytes / 34702 seconds) * 0.008 = 101.8 kbit/s

Winamp 5.62 FhG VBR 3: (456 149 050 bytes / 34702 seconds) * 0.008 = 105.2 kbit/s
(I guess, I must divide on original duration, right? If not, then bitrate will be 105.1 kbit/s )

So, that's all for now

This post has been edited by Steve Forte Rio: Jul 2 2011, 12:33

[IgorC]

Let's see how it compares with my results.

TVBR: --tvbr 46 --highest --samplerate keep. 93.62 kbps
CVBR: --cvbr 96 --highest --samplerate keep. 101.02 kbps
FhG VBR 3. 101.3 kbps
Nero -q 0.345. 101.89 kbps
Winamp CT is CBR ( hence doesn't need bitrate verification).



Excel table www.mediafire.com/?guujfe9qny2kb32

TVBR produces considerably lower bitrate. Windows version of Quicktime produces the same bitrate with --tvbr 45, 46 ...48. So there is no possibility to shift bitrates. While Mac version of QT can shift bitrates.
Unless some MAC user want to help here.
Other way we can include both modes CVBR and TVBR (but we should exclude Winamp's CT encoder then.)

Suggestions and opinions, please.

What are the "default" settings for AAC 96kbps on Itunes OSX and Itunes Windows?

iTunes CVBR equivalent : --cvbr 96 --normal --samplerate keep

[romor]

Windows version of Quicktime produces the same bitrate with --tvbr 45, 46 ...48.

I noticed the same yesterday while trying to tune switches, and at the end I figured you did fine job there

I tested more releases today also:

graph: http://i.imgur.com/i17Pq.png
boxplot: http://i.imgur.com/mNeRL.png
csv: http://pastebin.com/t1VtpNpY

from which my previous assumption seems wrong and your suggested parameters shows fine


highlight (149 tracks, 14 releases):

nero: 96
FgH: 93
QT-CVBR: 100
QT-TVBR: 94

[IgorC]

@romor
Thank you for your work.

What exact settings do you use for QT-CVBR and TVBR?

[romor]

suggested from post #76, of course
why do you ask?

This post has been edited by romor: Jul 2 2011, 22:59

[IgorC]

TVBR: --tvbr 46 --highest --samplerate keep
CVBR: --cvbr 96 --highest --samplerate keep

This way everybody will understand what we are talking about.



Current bitrate table:


This post has been edited by IgorC: Jul 2 2011, 23:11

[romor]

Also, it's interesting to see how flac and FgH bitrates correlate (in above set 0.8)
You and Steve did not provide data per track, maybe it can be noticed there as well?

This post has been edited by romor: Jul 2 2011, 23:20

[IgorC]

You and Steve did not provide data per track, maybe it can be noticed there as well?

I just didn't want to ask too much. But everybody can do some extra job as you've done.

This post has been edited by IgorC: Jul 2 2011, 23:34

[romor]

It's a same job, assuming flac version is source, but yeah, maybe not all find such things interesting

[IgorC]

I don't know why images aren't displayed properly.

Current bitrate table:
http://listening-test.blogspot.com/2011/07...rification.html




My particular bitrate table


This post has been edited by IgorC: Jul 3 2011, 00:29

[nao]

Windows version of Quicktime produces the same bitrate with --tvbr 45, 46 ...48. So there is no possibility to shift bitrates. While Mac version of QT can shift bitrates.

Mac version also produces the same bitrate. I guess you are confusing it with the --cvbr/abr/cbr switch, which accepts odd bitrates in the mac version.

[IgorC]

You're right. I've confused it with CVBR odd bitrates from old thread http://www.hydrogenaudio.org/forums/index....st&p=685749

Then we should make choice.
We can include both CVBR and TVBR. Single TVBR mode won't be enough representative due to its lower bitrate (-6.2%). But we should exclude Winamp CT encoder.
There were a lot of interest to compare (popular) iTunes CVBR and TVBR. http://www.hydrogenaudio.org/forums/index....st&p=682691

I know we will return to this question (CVBR at 100-101 kbps vs TVBR 95 kbps) again. There is no guaranty that TVBR at 95 kbps is any better than CVBR at 100-101 kbps.

New candidate list: Nero, TVBR, CVBR, Fhg.
Votes also show that comparison between CVBR and TVBR is inevitable http://listening-test.blogspot.com/2011/06...reparation.html

Objections?

This post has been edited by IgorC: Jul 3 2011, 12:13

[benski]

Duration of resultant files for all coders except Winamp FhG is identical to the original. Seems like FhG can't encode gaplessly - I've got +3 seconds totally for it.

Yes, unfortunately there is a bug that makes the new encoder sometimes not write the gapless data. I just fixed it. I'll post an updated DLL here shortly for everyone, and we should get the release out in about a week.

[IgorC]

Some previous public AAC tests included 5 codecs. So it's not impossible. http://listeningtests.t35.com/

It can be: Nero, QT CVBR, TVBR, Fhg, CT.

The bitrate verification will end on 7-8 July. This way there will be more time for sample selection 8-18 July and test will be started a few days earlier (20 July)

This post has been edited by IgorC: Jul 4 2011, 03:21

[Steve Forte Rio]

I'm for including of 5 codecs... It would be too unpleasant to exclude TVBR or CT AAC from the test.

Yes, unfortunately there is a bug that makes the new encoder sometimes not write the gapless data. I just fixed it. I'll post an updated DLL here shortly for everyone, and we should get the release out in about a week.

Thanks

This post has been edited by Steve Forte Rio: Jul 4 2011, 08:04