I've been wanting to re-encode my entire collection to AAC, since a) I now own a 5.5G iPod, and b) I've found quite a few problem samples with -V2 over the past few years, mostly relating to pre-echo, ringing, and sandy/coarse sounds, especially with acoustic jazz.
When testing out bitrates for a potential re-encode, I started out with 128kbps VBR in iTunes 7.0.2. I happened to have a remastered version of Herbie Hancock's 1965 post-bop classic, Maiden Voyage, laying next to my computer. After testing a few tracks, I had the feeling that something was not right - the instruments occasionally seemed to be "bleeding" into the other channel, much like crosstalk. When going back to compare against the ALAC file, I was right! QT AAC butchers this particular album in several places, but especially on the second track, where the tenor sax comes in and starts a solo. The drum kit, which is supposed to be slightly off center in the right channel, sounds like it is coming from the center. Worse still, there are metallic shimmers and watery sounds wavering back and forth between the channels. This is one of the worst artifacts I've ever heard with AAC, which usually is quite good to my ears, even at 128kbps.
Interestingly, this CD is an example of Rudy van Gelder's so-called "remastering," where he squashes his already dry and ambience-less recordings into a nearly monaural soundstage. No other encoder I've tried, though (Vorbis and LAME included) seems to have this same issue with imaging on this sample.
I've uploaded samples in the "uploads" section - all FLAC clips, 30 seconds each, with the AAC versions compressed with QT 7.1.3 at 128kbps VBR, 192kbps VBR, and 320kbps VBR. The problem is so obvious that ABX is just not needed here - you'd have to be deaf not t hear this. The 320kbps even has blips of problems at times, though it is mostly clean sounding. Looks like I'll have to stick with LAME for a little while longer.
EDIT: Actually, the clips were too big, so I cut them down to about 9 seconds each. You still get the point, though.
http://www.hydrogenaudio.org/forums/index....showtopic=41278 (http://www.hydrogenaudio.org/forums/index.php?showtopic=41278)
http://www.hydrogenaudio.org/forums/index....showtopic=41275 (http://www.hydrogenaudio.org/forums/index.php?showtopic=41275)
Looks like this issue is not new.
Earlier posters seemed to note issues with low frequencies, I found issues with high frequencies - looks like we now have all our bases covered
8/8 on original vs 128VBR. Looks like I may be retranscoding away from 128VBR in the near future
I wasn't immediately able to tell apart 192VBR from original through sighted listening so I didn't try an ABX test for that. It may be transparent to me. ER-4S headphones, AV710 sound card.
EDIT: Also confirmed with iTunes 7.0.2 encode on WinXP, so it is not merely a compiler/CPU bug of some kind.
I don't use iTune. But i tested your sample today because am bored. Encoded by me with iTune 7.0.2.16
320 kbps CBR (yes.. 320kbps)
I start with thunder from another post first.
foo_abx 1.3.1 report
foobar2000 v0.9.4.2
2007/04/13 19:02:29
File A: F:\Fraps\new download\thunder.flac
File B: E:\ALL MUSIC\test\1 thunder 2 320.m4a
19:02:29 : Test started.
19:03:00 : 01/01 50.0%
19:03:38 : 02/02 25.0%
19:03:57 : 03/03 12.5%
19:04:16 : 04/04 6.3%
19:04:49 : 05/05 3.1%
19:05:05 : 06/06 1.6%
19:05:20 : 07/07 0.8%
19:05:28 : 08/08 0.4%
19:05:56 : 09/09 0.2%
19:06:03 : 10/10 0.1%
19:06:39 : 11/11 0.0%
19:07:02 : 12/12 0.0%
19:07:22 : 13/13 0.0%
19:07:45 : 14/14 0.0%
19:07:57 : 15/15 0.0%
19:08:37 : 16/16 0.0%
19:08:44 : 17/17 0.0%
19:08:54 : 18/18 0.0%
19:09:09 : 19/19 0.0%
19:09:40 : 20/20 0.0%
19:09:42 : Test finished.
----------
Total: 20/20 (0.0%)
not very obvious like 128&192 kbps but it's still easy to ABX.
and with your sample.
foo_abx 1.3.1 report
foobar2000 v0.9.4.2
2007/04/13 19:22:00
File A: F:\Fraps\new download\Original.flac
File B: E:\ALL MUSIC\test\1 Original 320.m4a
19:22:00 : Test started.
19:22:44 : 01/01 50.0%
19:22:58 : 02/02 25.0%
19:23:13 : 03/03 12.5%
19:23:28 : 04/04 6.3%
19:25:31 : 05/05 3.1%
19:26:13 : 06/06 1.6%
19:26:24 : 07/07 0.8%
19:26:57 : 08/08 0.4%
19:27:03 : 09/09 0.2%
19:27:49 : 10/10 0.1%
19:28:28 : 11/11 0.0%
19:29:12 : 12/12 0.0%
19:29:15 : Test finished.
----------
Total: 12/12 (0.0%)
still.. easy. i stop at 12 because it's so easy. it's made me feel sorry for apple if i test futher.
I know this is a very old topic, but someone on ASR recently linked to it as an example of a problem with AAC encoding. Since I use AAC a lot I thought it was worth doing a test to check if this is something I should worry about.
Completed trials: 20
Number correct: 10 (50.0%)
Confidence that your results are better than chance: 0.41190147399902344 (41.190147399902344%)
Last played: A
Individual test results [choice/actual] from among 2 possible files:
Test #1 [ B / A ]
Test #2 [ A / A ]
Test #3 [ B / A ]
Test #4 [ B / B ]
Test #5 [ A / A ]
Test #6 [ A / A ]
Test #7 [ A / A ]
Test #8 [ B / B ]
Test #9 [ B / B ]
Test #10 [ B / A ]
Test #11 [ A / B ]
Test #12 [ B / A ]
Test #13 [ B / B ]
Test #14 [ A / A ]
Test #15 [ A / B ]
Test #16 [ B / B ]
Test #17 [ A / B ]
Test #18 [ A / B ]
Test #19 [ A / B ]
Test #20 [ B / A ]
Files in use:
A: /Users/charles/Public/QT AAC/Original.flac gain: 0.00 dBFS
B: /Users/charles/Public/QT AAC/The_Eye_Of_The_Hurricane_128_VBR.flac gain: 0.00 dBFS
I managed to convince myself I could hear a slight difference on the cymbals ... but this was clearly an illusion. Since I use AAC at around twice the bitrate tested it'll be just fine for my 55yr-old ears.
Sounds like my time using QTAAC I've found(uploaded) it needing 256kbps for Dark ambient, Since it mangles them. Yet no issues when trying them with 160kbps Vorbis/MPC & V3 Lame?.