IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Question to all VBR users - Bitrate limit, How much VBR is allowed?
C.R.Helmrich
post Jan 20 2010, 20:52
Post #1





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



Hi VBR users,

the results of 2008's public MP3 listening test (details here and here) show that for some very difficult material, a VBR encoder might spend more than 1.5 times the nominal bitrate (on average, the "Fatboy" item in that test was encoded at 208 kb/s, given a nominal bitrate of 128 kb/s).

I would like to ask the community if such a behaviour in VBR encoders is considered tolerable or is even encouraged (given the "difficulty" of the item in question).

Does any VBR user oppose bitrates higher than, say, 1.5 times the target bitrate?

Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
benski
post Jan 20 2010, 21:03
Post #2


Winamp Developer


Group: Developer
Posts: 670
Joined: 17-July 05
From: Brooklyn, NY
Member No.: 23375



In my opinion, the instantaneous bitrate should be allowed to vary up to the maximum allowed by the spec if that's what is required to meet the psychoacoustic model's demands. After all, VBR is meant to be "constant quality". If you look at a LAME -v5 encoding, you'll see it has frequent 320kbps frames even though the average bitrate is around 128kbps. That's 2.5x
Go to the top of the page
+Quote Post
jmcguckin
post Jan 20 2010, 21:13
Post #3





Group: Members
Posts: 52
Joined: 2-January 08
Member No.: 50051



as a QuickTime AAC True VBR user, I've gotten used to seeing average bitrates in my library go as high as 1.4x of the nominal bitrate that I use (and vice versa, as low as 50% of it)... and since the norm over my library tends to be within 20-25% of its average bitrate, the occasional outlier doesn't bother me at all.

plus, when watching the bitrate vary in foobar2000, it's usually pretty easy to identify the sounds that make the bitrate spike (particularly when the mix in one channel is very different from the other or if the file has a very large stereo width), and I tend to trust the encoder to make the proper decisions with bitrates, even if sometimes my instincts tell me it's going a little too far to either extreme.


--------------------
Archive- FLAC (-v 8)
Portable- QuickTime AAC (True VBR/-q 77)
Go to the top of the page
+Quote Post
kornchild2002
post Jan 20 2010, 22:58
Post #4





Group: Members
Posts: 2071
Joined: 8-April 05
From: Cincinnati, OH
Member No.: 21277



I don't appose having songs with much higher bitrates that the displayed "target" bitrate (I know it isn't a target bitrate as true VBR targets a constant quality). I know that a VBR encoder is doing its job if I end up with files that are 1.5X higher or lower than the displayed "terget" bitrate. I would rather have an encoder go up much higher than have it cutoff at a certain bitrate simply because it is too high. That was one of my concerns back in the day when using iTunes for constrained VBR AAC encoding. I was worried if the encoder was not selecting to use a higher or lower bitrate simply because it strayed too far away from the target bitrate (this was an actual target bitrate).
Go to the top of the page
+Quote Post
twostar
post Jan 20 2010, 23:16
Post #5





Group: Members
Posts: 487
Joined: 5-August 02
From: Manila
Member No.: 2939



I'm fine with the encoder going 1.5x times the target bitrate. If I wanted predictable file sizes, I'd just use CBR or ABR.
Go to the top of the page
+Quote Post
southisup
post Jan 20 2010, 23:31
Post #6





Group: Members
Posts: 251
Joined: 28-October 05
Member No.: 25414



I don't have any problem at all with 1.5 times target rate being exceeded occasionally. If it ever was decided to constrain bitrate range more rigidly then I hope it could be done by a "Strict mode" switch option.
Go to the top of the page
+Quote Post
halb27
post Jan 21 2010, 09:41
Post #7





Group: Members
Posts: 2425
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



IMO it's very welcome if an encoder on occasion chooses a bitrate much higher than average bitrate (average calculated for a large collection of tracks of various material).
First of all that's a local effect (thoough it can significantly increase average bitrate of an entire track, more so for track snippets used in listening tests).
With today's storage capacities I would welcome very much if within the various decision making processes the encoder uses a defensive choice for potentially
critical spots in the music.


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
psycho
post Jan 21 2010, 17:39
Post #8





Group: Members
Posts: 241
Joined: 14-October 05
Member No.: 25099



I also prefere if VBR uses more bps if it is needed. We've got ABR for predictable filesizes!


--------------------
lame -V 0
Go to the top of the page
+Quote Post
C.R.Helmrich
post Jan 22 2010, 18:48
Post #9





Group: Developer
Posts: 686
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



Thanks, everyone, for your comments!

I was asking because I noticed that, for the bit-hungriest item I could find, Fraunhofer's latest encoder goes up to about 1.5 times the target rate. But you convinced me that this is nothing I have to worry about smile.gif Of course, the instantaneous bitrate is allowed to be even higher, on the order of twice the target bitrate.


Chris


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
kornchild2002
post Jan 22 2010, 19:26
Post #10





Group: Members
Posts: 2071
Joined: 8-April 05
From: Cincinnati, OH
Member No.: 21277



Many people who listen to hard rock and metal are used to performance like that. For example, I have many tracks that Lame 3.98.2 ends up having an overall bitrate of 250-260kbps despite using the -V 2 command. Dimmu Borgir seems to make every lossy encoder (that I have used) end up with a higher bitrate than the "advertised" average for said settings. You can mess around with some of their songs to see how the FhG encoder performs (I only have access to CBR FhG mp3 encoding).
Go to the top of the page
+Quote Post
halb27
post Jan 22 2010, 19:43
Post #11





Group: Members
Posts: 2425
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



QUOTE (kornchild2002 @ Jan 22 2010, 19:26) *
... hard rock and metal ... Lame 3.98.2 ends up having an overall bitrate of 250-260kbps despite using the -V 2 command. ...

This is mp3's special sfb21 issue making VBR use a very high bitrate when there's a lot of high energy content beyond 16 kHz.
If you lowpass to 16 kHz or use the -Y switch you see VBR behavior without the sfb21 issue. Shouldn't be that extreme.


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
IgorC
post Jan 22 2010, 20:43
Post #12





Group: Members
Posts: 1553
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



QUOTE (halb27 @ Jan 22 2010, 15:43) *
QUOTE (kornchild2002 @ Jan 22 2010, 19:26) *
... hard rock and metal ... Lame 3.98.2 ends up having an overall bitrate of 250-260kbps despite using the -V 2 command. ...

This is mp3's special sfb21 issue making VBR use a very high bitrate when there's a lot of high energy content beyond 16 kHz.
If you lowpass to 16 kHz or use the -Y switch you see VBR behavior without the sfb21 issue. Shouldn't be that extreme.

-Y will also hurt quality at -V2 and hiher. -Y isn't solution for sfb21 issue. Developers of LAME disabled -y for reason at high bitrate. sfb21 is inevitable. I made a few internal tests before.

This post has been edited by IgorC: Jan 22 2010, 20:45
Go to the top of the page
+Quote Post
halb27
post Jan 22 2010, 20:51
Post #13





Group: Members
Posts: 2425
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



I only wanted to say that mp3's extremely high bitrate for many hard rock/metal tracks is primarily due to specific sfb21 issue and not due to general VBR behavior.

This post has been edited by halb27: Jan 22 2010, 20:52


--------------------
lame3100m -V1 --insane-factor 0.75
Go to the top of the page
+Quote Post
[JAZ]
post Jan 22 2010, 21:00
Post #14





Group: Members
Posts: 1772
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



Oh, please, IgorC. You're not a newcomer!.

Do you understand the problematic around the "sfb21" problem? Do you know where it puts the extra bitrate it uses? Your comment sounds like "if it can go to 11, everything else is bullshit".
Of course the output using the -Y switch differs from not using it, but this does not imply a degradation per-se.

We are talking about lossy codecs, and concretely about one designed more than 15 years ago. The -Y switch can help more than damage.
Just don't try to find the 11 in MP3.
Go to the top of the page
+Quote Post
kornchild2002
post Jan 22 2010, 21:01
Post #15





Group: Members
Posts: 2071
Joined: 8-April 05
From: Cincinnati, OH
Member No.: 21277



I understand, I was just giving an example of when VBR encoding ends up in files with higher than "advertised" bitrates. Not really going into specifics, just giving an example.

Edit: my comment was addressed to hal, not jaz (who wasn't there whenever I made my post hence why I didn't quote a statement).

This post has been edited by kornchild2002: Jan 22 2010, 21:58
Go to the top of the page
+Quote Post
greynol
post Jan 22 2010, 21:07
Post #16





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (IgorC @ Jan 22 2010, 11:43) *
-Y will also hurt quality at -V2 and hiher.

If this is about sound quality it is to be determined by ABX testing on an individual basis only. If you're speaking only on a conceptual level then this is fine (though may still be debatable), else it could be perceived by some as a TOS #8 violation.

This post has been edited by greynol: Jan 22 2010, 21:21


--------------------
Concern trolls: not a myth.
Go to the top of the page
+Quote Post
IgorC
post Jan 22 2010, 22:37
Post #17





Group: Members
Posts: 1553
Joined: 3-January 05
From: ARG/RUS
Member No.: 18803



JAZ,
I share all your standings. It's just a multiple misunderstood between me, Halb and you.
And it's my fault as I didn't express statement correctly. I didn't mean that -Y hurts quality by itself at the same bitrate as without -Y. I understand the concept of lossy encoder (tradeoff on size vs quality). -Y is fine for for every -V setting not only up to -V3.
It would be more logical to compare -V2 and -V1/0 -Y as bitrate will be comparable between both. It's right way to compare two lossy encoders,setting,etc..

But, yes, 11 is just so funny

This post has been edited by IgorC: Jan 22 2010, 22:47
Go to the top of the page
+Quote Post

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: 21st August 2014 - 14:12