Skip to main content
Topic: ABX for /mnt (Read 7880 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

ABX for /mnt

-

ABX for /mnt

Reply #1
A good example that even somtimes 320kbps can not be transparent.

LAME 3.97 -b 320

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.5.1
2008/03/27 21:00:39

File A: C:\Transcodes\MP3\One' Robots Dream 320 CBR.mp3
File B: E:\Downloads\satriani_sample.flac

21:00:39 : Test started.
21:01:07 : 01/01  50.0%
21:01:36 : 02/02  25.0%
21:02:19 : 03/03  12.5%
21:02:53 : 04/04  6.3%
21:03:26 : 04/05  18.8%
21:03:47 : 05/06  10.9%
21:04:28 : 06/07  6.3%
21:05:22 : 07/08  3.5%
21:05:58 : 08/09  2.0%
21:06:52 : 09/10  1.1%
21:07:32 : 10/11  0.6%
21:09:22 : 11/12  0.3%
21:12:22 : 12/13  0.2%
21:13:23 : Test finished.

----------
Total: 12/13 (0.2%)


A weird echo noise at around over 0:05 and the artifact is very noticable at the beginning of 0:06. This was very hard to ABX though, but am shocked that it was possbile to ABX at 320kbps CBR on a cheap Sound Blaster Audigy SE sound card and a cheap pair of Sennheiser HD 202 headphones, but there the best ever headphones I have ever owned though.

Note the file names or different because I transcoded from foobar2000 and it used the lossless file's metadata to name it.
"I never thought I'd see this much candy in one mission!"

ABX for /mnt

Reply #2
-

ABX for /mnt

Reply #3
What are we listening for?

(you haven't provided a link back to a discussion thread).

Cheers,
David.

ABX for /mnt

Reply #4
This is the only relevant thread I can find, and it makes no reference.  Perhaps this was agreed by PM.
I'm on a horse.

ABX for /mnt

Reply #5
...  on a cheap Sound Blaster Audigy SE sound card and a cheap pair of Sennheiser HD 202 headphones, but there the best ever headphones I have ever owned though. ...

Perhaps that's the reason you can. As has been mentioned many times before, equipment with a more linear response may disguise the problem better.
daefeatures.co.uk

ABX for /mnt

Reply #6
-

ABX for /mnt

Reply #7
What are we listening for?

(you haven't provided a link back to a discussion thread).

Cheers,
David.

There is a odd echo artifact on the drums at 0:06.
"I never thought I'd see this much candy in one mission!"

ABX for /mnt

Reply #8
-

ABX for /mnt

Reply #9
I had this bitrate dilemma for a while. But gradually I got off the very high bitrate train. Still trying to find strong reasons to abandon V3..

Percieved quality : V3 - 8.5, V2 - 8.7, V0 - 9.1, Insane - 9.3 or so. Bitrate eventually goes crazy on higher settings while still not fully transparent under close abxing. Sbf21 is an issue with HF heavy music , the coding above 16khz results in overcoding *below* 16khz (bloat). -Y is experimental and I don't want to rely too much on hack commandlines. I tried a few times -Y vs nomal v2 etc.. I can't hear what is missing.. Who can ?? Dibrom said that even if you could the quality above 16khz will be terrible.

I think for possible higher quality at 160..190k  with very high freq MPC, Vorbis and maybe AAC are more efficient options. Since 128..160k is near transparent for many then a good codec should give transparent result with 160..190 on most samples. For > 320k  bitrates wavpack lossy , lossywav will be better and 384k isn't far from 320k mp3.
wavpack 4.8 -b3x6c

ABX for /mnt

Reply #10
I had this bitrate dilemma for a while. But gradually I got off the very high bitrate train. Still trying to find strong reasons to abandon V3..

Percieved quality : V3 - 8.5, V2 - 8.7, V0 - 9.1, Insane - 9.3 or so. Bitrate eventually goes crazy on higher settings while still not fully transparent under close abxing. Sbf21 is an issue with HF heavy music , the coding above 16khz results in overcoding *below* 16khz (bloat). -Y is experimental and I don't want to rely too much on hack commandlines. I tried a few times -Y vs nomal v2 etc.. I can't hear what is missing.. Who can ?? Dibrom said that even if you could the quality above 16khz will be terrible.

I think for possible higher quality at 160..190k  with very high freq MPC, Vorbis and maybe AAC are more efficient options. Since 128..160k is near transparent for many then a good codec should give transparent result with 160..190 on most samples. For > 320k  bitrates wavpack lossy , lossywav will be better and 384k isn't far from 320k mp3.

I decided to ditch the idea using V0 since its will not be 100% transparent to me on some rare tracks, sometimes that is not even transparent agaisnt pre-echo artifacts but effective when -vbr-new is performing worse then vbr-old (very rare), but most of time is effective but I don't like the bitrates I would get from V0 thanks to the sbf21 issue and all must of my mp3s would up as 280 - 302 at V0 since I listen to alot of metal or metalcore since those generes have alot hi-freqs and 320 CBR is a overkill but sometimes it is not enough on some very very rare samples or if you are using a really bad encoder such the first version of L3ENC or Blade. About 98% of my music collection should be transparent on V2 anyway and I got a good balance of sound quality and file size on V2 unlike I would on V0 or 320kbps.
"I never thought I'd see this much candy in one mission!"

ABX for /mnt

Reply #11
...I don't like the bitrates I would get from V0 thanks to the sbf21 issue ... since I listen to alot of metal ... and I got a good balance of sound quality and file size on V2 ....

As shadowking said you don't have to worry about sfb21 if you explicitly use -Y which is defaulted for -V3 and below.
Chance is with metal you arrive at a lower bitrate when using -V0 -Y than when using plain -V2.
Other than that it's always personal taste where to put the sweetspot: -V3, -V2 -Y, -V1 -Y, -V0 -Y all have their merits.
From my personal experience average bitrate with -V0 -Y  (3.98b6) is around 225 kbps when used on those tracks from my collection I decided to encode to mp3.
lame3995o -Q1.7
opus --bitrate 140

ABX for /mnt

Reply #12
Another option might be --lowpass 17
Slightly less effective than a 16k lowpass but it does save bits while encoding stuff in 16.5 ~ 17 khz (more than enough IMO). I would use for V2 ~ V0 range.
wavpack 4.8 -b3x6c

ABX for /mnt

Reply #13
-

ABX for /mnt

Reply #14
Yeah I still think V4 maybe the best compromise, but V3 is still very economical across all genres and there are some quality gains (less distorted sound - ringing). I noticed it in lower volume stuff, intros, solo instrument etc. V5 is definately not transparent , sometimes ringing badly. V4 reduces this and its mostly gone around V3 / V2..

Anyway, I set my limits because in reality V5 is already quite good for most use , if there are obvious problems with v3 , v2 won't be much better but can be much fatter.
wavpack 4.8 -b3x6c

ABX for /mnt

Reply #15
I decided to ditch the idea using V0 since its will not be 100% transparent to me on some rare tracks, sometimes that is not even transparent agaisnt pre-echo artifacts but effective when -vbr-new is performing worse then vbr-old (very rare), but most of time is effective but I don't like the bitrates I would get from V0 thanks to the sbf21 issue and all must of my mp3s would up as 280 - 302 at V0 since I listen to alot of metal or metalcore since those generes have alot hi-freqs and 320 CBR is a overkill but sometimes it is not enough on some very very rare samples or if you are using a really bad encoder such the first version of L3ENC or Blade.



Whew.  Punctuation can be your friend, you know. 

ABX for /mnt

Reply #16
I listen to alot of metal or metalcore ...
You won't need to worry about stuff above 16kHz for long then

Sorry - that's a stereotype - maybe you listen to it quietly?

If you listen to it at the volume I hear other people using on the tube, and do it for long enough each day, then you'll train your ears to ignore higher frequencies quickly enough.


On a serious note, it does sound like mp3 may not be for you.


EDIT: Can't ABX V2.

Cheers,
David.

ABX for /mnt

Reply #17
Using --lowpass 17 on a few of those loud metalish tracks managed to save around 10 % with V2 ~ V0 (20~30 kbit saving) .. Personally I like a -Y type filter better.
wavpack 4.8 -b3x6c

ABX for /mnt

Reply #18
Using --lowpass 17 on a few of those loud metalish tracks managed to save around 10 % with V2 ~ V0 (20~30 kbit saving) .. Personally I like a -Y type filter better.

Better using V3 since it uses the -Y switch, since V3 does not deal with very high freqs it could possibly outperform V2 on some situations, if the listener can not hear over 17Khz or sound equipment that has poor freq response.
"I never thought I'd see this much candy in one mission!"

 
SimplePortal 1.0.0 RC1 © 2008-2019