Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Personal Multiformat ABC/HR test at 96 kbit/s (Read 29357 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Personal Multiformat ABC/HR test at 96 kbit/s

Codecs :
High anchor  Lame 3.98b1 --vbr-new -V 5
Low anchor Lame 3.98b1 --abr 96
Nero 1.0.7.0 VBR -q 0.34 (approx 96 kbps)
Itunes 7.1.1.5  96 kbit/s VBR
WMA10pro 96 kbit/s CBR 1pass (dbpower converter 12.1)
Lancer Vorbis Aotuv (last oggdropXP nov 10 2006) -q 2.00 (approx 96 kbps)

Decoders:
Nero, itunes, Lancer, lame were decoded by foobar 0.9.4.3
wma10pro by dbprower converter 12.1

Headphones Sennheiser HD 447 and soundcard Audigy SE 24/96

Results:
http://rapidshare.com/files/34474605/96kbps.xls.html

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #1
Thanks for your efforts! Just one suggestion, put the anchors on the left/right ends of the table. Less confusing

But thanks!

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #2
Thanks for your efforts! Just one suggestion, put the anchors on the left/right ends of the table. Less confusing

But thanks!


Thanks for suggestion.
I added more samples to test.

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #3
I added more samples. Now it's 50. New table http://rapidshare.com/files/34474605/96kbps.xls.html

iTunes surprised. It cuts more frecuencies than other encoders however has a higher score.

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #4
Your excel file made it easy to cut and paste to the statistical webalyzer at my site:

http://ff123.net/friedman/stats.html

All tied except for the low anchor.  There is the suggestion (though not statistically validated) that you prefer itunes and Lancer over Nero.


Code: [Select]
FRIEDMAN version 1.24 (Jan 17, 2002) http://ff123.net/
Tukey HSD analysis

Number of listeners: 50
Critical significance:  0.05
Tukey's HSD:   0.335

Means:

itunes   Lancer   Lame_v5  wma10pro Nero     Lame_abr
  4.60     4.58     4.46     4.45     4.33     3.20  

-------------------------- Difference Matrix --------------------------

         Lancer   Lame_v5  wma10pro Nero     Lame_abr
itunes     0.014    0.136    0.146    0.264    1.396*
Lancer              0.122    0.132    0.250    1.382*
Lame_v5                      0.010    0.128    1.260*
wma10pro                              0.118    1.250*
Nero                                           1.132*
-----------------------------------------------------------------------

itunes is better than Lame_abr
Lancer is better than Lame_abr
Lame_v5 is better than Lame_abr
wma10pro is better than Lame_abr
Nero is better than Lame_abr

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #5
There is a trouble. The number format is x,xx in Excel and not x.xx as here http://ff123.net/friedman/stats.html

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #6
Look at that, most of the formats are matching or slightly better in quality than the high anchor!  I haven't done any ABX to verify but I suspect my results would be similar.  I'm content with 80 kbps Vorbis on my portable after all.  This would be interesting to see again with the same high anchor but testing for 80 kbps transparency.

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #7
My try.

Encoders
96kbps contenders
aoTuV beta5 (q 2)
iTunes v7.3.1.3 (96k VBR)
Nero AAC Enc v1.1.34.0 (q 0.35)

128kbps contenders
LAME v3.97 (-V5 --vbr-new)
LAME v3.98b4 (-V5)

Samples
All samples comes from Sebastian's choice on 64kbps multi-format listening test. But I couldn't finish every one of them, because it's too hard for this comparison or unfamiliar to me.

Results
Code: [Select]
sample n    aoTuV    iTunes    L397    L398    Nero
sample 01    3.50    4.20    3.80    2.70    4.30
sample 02    3.50    4.30    4.20    4.20    3.40
sample 03    3.50    4.10    4.30    4.30    4.00
sample 04    3.90    4.10    3.50    3.70    4.20
sample 05    3.00    4.00    5.00    4.00    3.70
sample 06    3.00    5.00    5.00    5.00    5.00
sample 07
sample 08    3.50    5.00    5.00    5.00    5.00
sample 09    3.20    3.50    4.00    2.80    4.20
sample 10    4.00    4.50    4.50    4.00    5.00
sample 11
sample 12    3.90    3.70    3.20    3.90    4.00
sample 13    3.30    4.00    3.50    2.80    4.00
sample 14    3.20    4.00    4.20    3.70    4.00
sample 15    3.50    4.30    4.50    5.00    4.00
sample 16
sample 17    4.00    4.10    3.70    3.20    4.20
sample 18    3.40    4.20    3.70    3.90    4.20
------------------------------------------------------------
Overall      3.49    4.2    4.14    3.88    4.21


AACs on top, followed by LAME 3.97, then 3.98. aoTuV loses.

Overall quality at this bitrate isn't impressive. I was rarely required to ABX. But AACs and 128kbps MP3 perform well with orchestral samples (No. 6 and 8).

Quote
FRIEDMAN version 1.24 (Jan 17, 2002) http://ff123.net/
Tukey HSD analysis

Number of listeners: 15
Critical significance:  0.05
Tukey's HSD:  0.458

Means:

Nero    iTunes  LAME397  LAME398  aoTuV   
  4.21    4.20    4.14    3.88    3.49 

-------------------------- Difference Matrix --------------------------

        iTunes  LAME397  LAME398  aoTuV   
Nero      0.013    0.073    0.333    0.720*
iTunes              0.060    0.320    0.707*
LAME397                      0.260    0.647*
LAME398                              0.387 
-----------------------------------------------------------------------

Nero is better than aoTuV
iTunes is better than aoTuV
LAME397 is better than aoTuV

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #8
haregoo
Thanks for share the results

Speaking of my test also I think Itunes 96 kbit/s VBR mode doesn't produce 96 kbit/s. For a lot of albums with enough silence itunes still has 105 kbit/s  while Vorbis -q2 and Nero 1.0.7.0 produce something like 89-90 kbits.On average Vorbis and Nero respect with their VBR mode 96 kbit/s bitrate while Itunes always ends on 105-110 kbit/s.

wouldn't be more fair to shift bitrate? something like:
Vorbis -q 2.5 (Instead of -q 2.0)
Nero 1.0.7.0 -q 0.36
Nero 1.1.34 -q 0.34 / -q0.35(?)

But it would be test for 105 kbit/s . Not 96 kbit/s.

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #9
What you can expect with Nero 1.1.34 is:

-q 0.34  ->  92kbps
-q 0.35  ->  96kbps
-q 0.37  ->  103kbps
-q 0.38  ->  107kbps

So for test at 105kbps one should be using -q 0.375

 

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #10
Speaking of my test also I think Itunes 96 kbit/s VBR mode doesn't produce 96 kbit/s. For a lot of albums with enough silence itunes still has 105 kbit/s  while Vorbis -q2 and Nero 1.0.7.0 produce something like 89-90 kbits.On average Vorbis and Nero respect with their VBR mode 96 kbit/s bitrate while Itunes always ends on 105-110 kbit/s.

Perhaps iTunes AAC is more like CBR unlike quality-based VBR encoder including aoTuV and Nero VBR setting. iTunes's one doesn't save bitrate for quieter simple part of music, thus most of music will end up with 96-105kbps. As far as these above samples are concerned, I could get fair bitrate table.

Personal Multiformat ABC/HR test at 96 kbit/s

Reply #11

Speaking of my test also I think Itunes 96 kbit/s VBR mode doesn't produce 96 kbit/s. For a lot of albums with enough silence itunes still has 105 kbit/s  while Vorbis -q2 and Nero 1.0.7.0 produce something like 89-90 kbits.On average Vorbis and Nero respect with their VBR mode 96 kbit/s bitrate while Itunes always ends on 105-110 kbit/s.

Perhaps iTunes AAC is more like CBR unlike quality-based VBR encoder including aoTuV and Nero VBR setting. iTunes's one doesn't save bitrate for quieter simple part of music, thus most of music will end up with 96-105kbps. As far as these above samples are concerned, I could get fair bitrate table.


I was told that Itunes VBR can only select larger frames and in this case will never have a frame at less than 96. In this respect it does not sound like a true VBR implementation, more like enhanced CBR.