IPB

Welcome Guest ( Log In | Register )

3 Pages V  < 1 2 3  
Reply to this topicStart new topic
Quicktime TVBR vs Nero AAC
Larson
post Feb 25 2011, 13:36
Post #51





Group: Members
Posts: 131
Joined: 27-March 09
Member No.: 68422



Mac OS X Lion beta has Quicktime 7.7.1, have to see if there are major differences in the AAC encoding. If someone is testing it take a look
Go to the top of the page
+Quote Post
/mnt
post Feb 25 2011, 21:51
Post #52





Group: Members
Posts: 697
Joined: 22-April 06
Member No.: 29877



I find QuickTime to be pretty weak against sharp transients, for example The Robot by Kraftwerk has lots of pre-echo on lower bitrates with QuickTime then with Nero AAC. Sometimes the pre-echo artifacts produced by QuickTime is just as bad as LAME's.

For example i just managed to ABX an random track at tvbr 90, due to smearing at 0:08:
CODE
foo_abx 1.3.4 report
foobar2000 v1.1.2
2011/02/25 20:37:48

File A: C:\Temp\Devo - New Traditionalists\1.01. Through Being Cool.m4a
File B: I:\Music\Devo - New Traditionalists\01. Through Being Cool.flac

20:37:48 : Test started.
20:38:14 : 01/01 50.0%
20:38:24 : 02/02 25.0%
20:38:38 : 03/03 12.5%
20:38:54 : 04/04 6.3%
20:38:59 : 05/05 3.1%
20:39:05 : 06/06 1.6%
20:39:11 : 07/07 0.8%
20:39:15 : 08/08 0.4%
20:39:30 : 09/09 0.2%
20:39:37 : 10/10 0.1%
20:39:46 : 11/11 0.0%
20:39:52 : 12/12 0.0%
20:39:57 : 13/13 0.0%
20:40:02 : 14/14 0.0%
20:40:10 : 15/15 0.0%
20:40:12 : Test finished.

----------
Total: 15/15 (0.0%)


--------------------
"I never thought I'd see this much candy in one mission!"
Go to the top of the page
+Quote Post
Steve Forte Rio
post Feb 25 2011, 23:01
Post #53





Group: Members
Posts: 432
Joined: 4-October 08
From: Ukraine
Member No.: 59301



/mnt, have you tried this sample with qt aac?

beats.tak


If not - try to encode it with qaac -V 127 -q2 and listen smile.gif

It looks like quicktime's vbr algoritm problem...

This post has been edited by Steve Forte Rio: Feb 25 2011, 23:02
Go to the top of the page
+Quote Post
/mnt
post Feb 26 2011, 02:09
Post #54





Group: Members
Posts: 697
Joined: 22-April 06
Member No.: 29877



QUOTE (Steve Forte Rio @ Feb 25 2011, 23:01) *
/mnt, have you tried this sample with qt aac?

beats.tak


If not - try to encode it with qaac -V 127 -q2 and listen smile.gif

It looks like quicktime's vbr algoritm problem...


The VBR on QuickTime seems to be buggy:

CODE
foo_abx 1.3.4 report
foobar2000 v1.1.4
2011/02/26 00:50:11

File A: C:\Downloads\beats.tak
File B: C:\Temp\beats tvbr 127.m4a

00:50:11 : Test started.
00:50:18 : 01/01 50.0%
00:50:21 : 02/02 25.0%
00:50:27 : 03/03 12.5%
00:50:30 : 04/04 6.3%
00:50:34 : 05/05 3.1%
00:50:38 : 06/06 1.6%
00:50:44 : 07/07 0.8%
00:50:49 : 08/08 0.4%
00:50:54 : 09/09 0.2%
00:51:02 : 10/10 0.1%
00:51:07 : 11/11 0.0%
00:51:12 : 12/12 0.0%
00:51:27 : 13/13 0.0%
00:51:29 : 14/14 0.0%
00:51:35 : 15/15 0.0%
00:51:40 : 16/16 0.0%
00:51:51 : Test finished.

----------
Total: 16/16 (0.0%)

Very obivious artifacts through out.

Going lower to tvbr 90 shows something interesting.

CODE
foo_abx 1.3.4 report
foobar2000 v1.1.4
2011/02/26 00:53:05

File A: C:\Temp\beats tvbr 90.m4a
File B: C:\Temp\beats tvbr 127.m4a

00:53:05 : Test started.
00:53:12 : 01/01 50.0%
00:53:16 : 02/02 25.0%
00:53:20 : 03/03 12.5%
00:53:26 : 04/04 6.3%
00:53:29 : 05/05 3.1%
00:53:32 : 06/06 1.6%
00:53:36 : 07/07 0.8%
00:53:41 : 08/08 0.4%
00:53:45 : 09/09 0.2%
00:53:49 : 10/10 0.1%
00:53:53 : 11/11 0.0%
00:53:56 : 12/12 0.0%
00:53:59 : 13/13 0.0%
00:54:03 : 14/14 0.0%
00:54:08 : 15/15 0.0%
00:54:12 : 16/16 0.0%
00:54:16 : Test finished.

----------
Total: 16/16 (0.0%)


It sounds fine to me at 90, but at 127 it has horrid pre-echo. Bitrates are similar aswell, the tvbr 90 encode is at 62kbps and the 127 encode is at 65kbps.

This post has been edited by /mnt: Feb 26 2011, 02:11


--------------------
"I never thought I'd see this much candy in one mission!"
Go to the top of the page
+Quote Post
TechVsLife
post Feb 28 2011, 20:25
Post #55





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



QUOTE (/mnt @ Feb 25 2011, 20:09) *
It sounds fine to me at 90, but at 127 it has horrid pre-echo. Bitrates are similar aswell, the tvbr 90 encode is at 62kbps and the 127 encode is at 65kbps.


Two questions:
1. Has this been reported to Apple as a bug (--I don't know if apple actually responds)?
2. I assume this bug is only on the tvbr side--no such bug detected using cvbr at 256 (the itunes plus setting) or higher? (or does one need to use cbr to avoid it).

thank you.

This post has been edited by TechVsLife: Feb 28 2011, 20:48
Go to the top of the page
+Quote Post
TechVsLife
post Mar 3 2011, 19:35
Post #56





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



I also have a more general question, if this bug affects quicktime cVBR 320 and tVBR 127, but not quicktime CBR 320, would this suggest (when sticking with aac super-high bitrates) that CBR 320 is better as an archival format? Assuming no flaws, VBR should in principle always be better; but given that flaws exist, is there reason to think CBR would likely be safer, in these very high bitrate ranges? I know there have been at least a couple of instances before (of a VBR flaw not appearing in equivalent CBR, at moderately high bitrates) involving lame, but perhaps that's all there is to it.
Go to the top of the page
+Quote Post
C.R.Helmrich
post Mar 4 2011, 11:01
Post #57





Group: Developer
Posts: 682
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



If space doesn't matter to you, then yes, CBR 320 should be "safer" than VBR 320. You'll probably overcode a lot of tracks, but you can be relatively sure that nothing is undercoded. As I wrote in another thread, writing a good VBR algorithm is more complicated than one might think (danger of undercoding).

At 320 it doesn't matter whether you use MP3 or AAC. I prefer AAC for technical reasons (it was developed as an "advanced" version of MP3).

By the way, I ran some HA test items through iTunes 10.2, which was released yesterday, at 128 CVBR and got different results than for the latest 9.x iTunes version. Most notably, some pre-echos seem to have been solved, which should be interesting for /mnt, in particular.

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
SamDeRe81
post Mar 4 2011, 15:48
Post #58





Group: Members
Posts: 70
Joined: 24-November 10
Member No.: 85992



=/ I want to re-encode my stuff but I think I'll use Nero, this quicktime sounds not so great =9 Can I ask you Mr. C.R.Helmrich what do you think a good vbr bitrate is? I was thinking 160?
Go to the top of the page
+Quote Post
Larson
post Mar 4 2011, 16:47
Post #59





Group: Members
Posts: 131
Joined: 27-March 09
Member No.: 68422



QUOTE (SamDeRe81 @ Mar 4 2011, 15:48) *
=/ I want to re-encode my stuff but I think I'll use Nero, this quicktime sounds not so great =9 Can I ask you Mr. C.R.Helmrich what do you think a good vbr bitrate is? I was thinking 160?


you have to use your ears and abx to judge what is good enough for you, you may be surprised by the results you'll get.
Go to the top of the page
+Quote Post
C.R.Helmrich
post Mar 4 2011, 20:27
Post #60





Group: Developer
Posts: 682
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



I agree with Larson. Start with some ABX test of, say, 128 kbps. This might already be so close to transparency for you that it suffices. It certainly is for me for "casual listening" (for "serious listening" I use FLAC).

By the way, in a private blind MUSHRA test nero vs. iTunes which I ran last year at ~128 kbps VBR, iTunes won by a small but significant margin. So if you decide to use something near that bit rate, you should give iTunes 10.2 a try in order to get a "complete picture".

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
TechVsLife
post Mar 4 2011, 21:58
Post #61





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



QUOTE (C.R.Helmrich @ Mar 4 2011, 05:01) *
If space doesn't matter to you, then yes, CBR 320 should be "safer" than VBR 320. You'll probably overcode a lot of tracks, but you can be relatively sure that nothing is undercoded. As I wrote in another thread, writing a good VBR algorithm is more complicated than one might think (danger of undercoding).

Thank you, that's very helpful. This test result, where the higher vbr rate (the maximum!) is doing a lot worse than the lower vbr rate, opens one's eyes to some of the difficulties in vbr encoding. Also I'd venture to guess that at super-high bitrates VBR may not have offsetting advantages (that tvbr at max bitrate is increasing risk of glaring undercodes compared to cbr 320 while any vbr overshooting of the fixed cbr rate is without benefit at 320)--that's a different situation than 128 or 192 bitrates (where vbr could be "safer," i.e. less likely to produce glaring artifacts than cbr comparable bitrate range). (I know it largely defeats the purpose of lossy to use 320, but I want to stick with only one version of every piece in a good near universal format.) I wonder if aac cbr 320 has bested aac cbr 256 in any samples, but I assume, unlike tvbr, it will never be worse.

QUOTE (C.R.Helmrich @ Mar 4 2011, 05:01) *
By the way, I ran some HA test items through iTunes 10.2, which was released yesterday, at 128 CVBR and got different results than for the latest 9.x iTunes version. Most notably, some pre-echos seem to have been solved, which should be interesting for /mnt, in particular.

I think 10.2 itunes, unlike 9.x, has the same quicktime as 10.1; both had qt 7.6.9--though 10.2 also shows "1680.9" which might be new. I don't know what quicktime /mnt or Steve Forte Rio used; it would be helpful to always post the version.

This post has been edited by TechVsLife: Mar 4 2011, 22:26
Go to the top of the page
+Quote Post
TechVsLife
post Mar 4 2011, 22:06
Post #62





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



QUOTE (C.R.Helmrich @ Mar 4 2011, 14:27) *
(for "serious listening" I use FLAC).
Chris

Have you been able to abx between aac cbr 320 and flac? (I'd assume you do it for other reasons; perhaps you're too aware of the "guts" of lossy codecs.)
Go to the top of the page
+Quote Post
C.R.Helmrich
post Mar 5 2011, 01:26
Post #63





Group: Developer
Posts: 682
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (TechVsLife @ Mar 4 2011, 23:06) *
QUOTE (C.R.Helmrich @ Mar 4 2011, 14:27) *
(for "serious listening" I use FLAC).
Chris

Have you been able to abx between aac cbr 320 and flac? (I'd assume you do it for other reasons; perhaps you're too aware of the "guts" of lossy codecs.)

By "serious listening" I simply meant listening to what I bought, which is CD audio. Having said that, yes, IIRC I was able to ABX either the eig sample or one of /mnt's Kraftwerk samples from here.

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
Wombat
post Mar 5 2011, 01:57
Post #64





Group: Members
Posts: 950
Joined: 7-October 01
Member No.: 235



I am able to abx "some" of the stuff /nmt brings up here and there. I never were good at finding pre-echo samples.
But it must be said that /nmt shouldn´t be a guide for anyone to adjust his settings to. I bet that many things /nmt finds to have problems is a non-issue for most human beings smile.gif

God must have blessed him with some fine set of pre-echo ears!

This post has been edited by Wombat: Mar 5 2011, 02:00
Go to the top of the page
+Quote Post
TechVsLife
post Mar 5 2011, 20:44
Post #65





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



QUOTE (Wombat @ Mar 4 2011, 19:57) *
But it must be said that /mnt shouldn´t be a guide for anyone to adjust his settings to. I bet that many things /mnt finds to have problems is a non-issue for most human beings :)


The concern in /mnt's finding is this:
"It sounds fine to me at 90, but at 127 it has horrid pre-echo."
The issue is not only that the artifact is very strong (horrid) to him, but especially that it is very strong at the max. vbr setting (127) while fine at a much lower setting (90)--with a caveat that I don't know if quicktime 7.6.9 was used. There are two standards: what I would detect (few if any artifacts) and what other people listening would detect in good conditions. The latter is hard to put a percentage on, but I'd like to have that also as close to zero as possible, while keeping only a "universal" file (i.e. aac or mp3). Under these circumstances, and where apparently bugs or less-than-perfectly-tuned algorithms still exist, CBR very high bitrate (256 or 320) might be safer than VBR (even vbr at equivalent bitrates). --For people who don't mind the hassle of multiple versions, of course keeping lossless and lossy versions is better.

This post has been edited by TechVsLife: Mar 5 2011, 20:46
Go to the top of the page
+Quote Post
TechVsLife
post Mar 6 2011, 05:23
Post #66





Group: Members
Posts: 195
Joined: 29-May 07
Member No.: 43837



I also wonder if (like CBR 256) quicktime ABR 256 or higher is free of this failure, that the tvbr and cvbr modes have.
Go to the top of the page
+Quote Post
nu774
post Apr 22 2011, 14:12
Post #67





Group: Developer
Posts: 477
Joined: 22-November 10
From: Japan
Member No.: 85902



QUOTE (Steve Forte Rio @ Feb 26 2011, 07:01) *
beats.tak
If not - try to encode it with qaac -V 127 -q2 and listen smile.gif

It looks like quicktime's vbr algoritm problem...

Hi, I tried this one.
Indeed, with strategies other than CBR, QuickTime seems to give too low bitrate for this sample. As a result, very obvious artifacts are produced.

I sent bug report to Apple about this, and they recently sent me back the mail.
They say that this issue has been addressed in Mac OS X Lion Developer Preview Build 11A419.

Can anyone here try the developer preview? I don't have environments to try it.
Go to the top of the page
+Quote Post
Brand
post Dec 4 2011, 14:08
Post #68





Group: Members
Posts: 312
Joined: 27-November 09
Member No.: 75355



I did some quick ABXing and I wasn't too impressed with Nero AAC. It performed a little bit worse than LAME (3.99) at the same bitrate/file size.

But I only tested with one sample, uploaded here (the 'original' one).

I'm wondering if Quicktime AAC performs better, at least for this sample. I don't want to install Quicktime and learn all the settings, so could someone upload a properly encoded QT AAC version? I'm looking for a bitrate similar to that of V3, so I think the file size should be somewhere around 600-650 KB (I wouldn't mind multiple files of different sizes).
Go to the top of the page
+Quote Post
ShotCaller
post Dec 4 2011, 21:10
Post #69





Group: Members
Posts: 34
Joined: 8-August 11
Member No.: 92854



This is QT TVBR q82, which is the closest in bitrate and filesize as -V 3 (LAME 3.98.4). I hope this helps you.

http://www.multiupload.com/JV4I2MBNUO
Go to the top of the page
+Quote Post
Brand
post Dec 5 2011, 00:44
Post #70





Group: Members
Posts: 312
Joined: 27-November 09
Member No.: 75355



Thanks!

The file is smaller than I expected, but it looks like 3.98 produces slightly smaller (and IMO slightly worse sounding) files at V3 compared to 3.99, since with the latter the V3 conversion results in a 622KB file [for this sample].

That example is still somewhat easily ABXable to me, although I did get the impression that it's better than Nero AAC, considering the file size. So if you don't mind could you post an example of a slightly higher quality? Something around 650KB file size for that sample.
Go to the top of the page
+Quote Post
ShotCaller
post Dec 5 2011, 01:06
Post #71





Group: Members
Posts: 34
Joined: 8-August 11
Member No.: 92854



Sure, this is q91. Bitrate is 178kbps average and size is 632kb.
Go to the top of the page
+Quote Post
Brand
post Dec 5 2011, 11:17
Post #72





Group: Members
Posts: 312
Joined: 27-November 09
Member No.: 75355



Thanks. I could still ABX that without too much difficulty. For a similar sized Nero AAC sample I used Q0.54 (~195kbps) in Foobar and got similar acoustic results. I can't honestly say I'd give a clear advantage to either.

LAME (3.99.3) seemed to perform better than both in terms of transparency. With V3 (which produced a slightly smaller file) I had more trouble ABXing.

It's only a short test with a single sample, so far from conclusive. I could also be biased since MP3 is a more convenient format for me (FolderPlay on my phone can't play AACs in MP4 files).

This post has been edited by Brand: Dec 5 2011, 11:17
Go to the top of the page
+Quote Post
eahm
post Oct 24 2012, 19:21
Post #73





Group: Members
Posts: 888
Joined: 11-February 12
Member No.: 97076



QUOTE (Brand @ Dec 5 2011, 03:17) *
Thanks. I could still ABX that without too much difficulty. For a similar sized Nero AAC sample I used Q0.54 (~195kbps) in Foobar and got similar acoustic results. I can't honestly say I'd give a clear advantage to either.

LAME (3.99.3) seemed to perform better than both in terms of transparency. With V3 (which produced a slightly smaller file) I had more trouble ABXing.

It's only a short test with a single sample, so far from conclusive. I could also be biased since MP3 is a more convenient format for me (FolderPlay on my phone can't play AACs in MP4 files).

Wow what? Old post but we need to see the ABX log.

This post has been edited by eahm: Oct 24 2012, 19:21
Go to the top of the page
+Quote Post
Brand
post Oct 25 2012, 11:38
Post #74





Group: Members
Posts: 312
Joined: 27-November 09
Member No.: 75355



The ABX log regarding what specifically?

Anyway, for MP3 vs AAC comparisons, I remember I could actually ABX V3 just as easily later, so you can disregard that.

But where AAC has a clear advantage over MP3 is not in the V3 bitrates, but around 120-130kbps (especially with the FhG encoder, which IMO is one of the best).
Go to the top of the page
+Quote Post

3 Pages V  < 1 2 3
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 24th April 2014 - 13:23