IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
MP3 Bit Reservoir, Did I understand it right ?
krafty
post Aug 30 2012, 02:53
Post #1





Group: Members
Posts: 266
Joined: 20-March 10
Member No.: 79175



Ok, so I've read some posts and learned a bit about the principles, but, something is still not clear...

- In V0, there is room for frames to go higher than 320kbps while sparing these bits on the low bitrate frames? (It can go a bit higher than 320 kbps limit)

- In 320 CBR, that doesn't happen, because the frame is literally stuck and can't got higher than 320 kbps, and thus it wastes on the frames that were with low bitrate. (It cannot go higher the 320kbps limit)

Is it still true that VBR would not be so consistent/precise as 320 in terms of choosing the right bitrate for a particular frame: eg: a frame needs 256 k but it "guesses" and calculates 250 k and then it was not truly consistent?

Thanks for answers.

This post has been edited by krafty: Aug 30 2012, 02:55
Go to the top of the page
+Quote Post
kode54
post Aug 30 2012, 03:30
Post #2





Group: Admin
Posts: 4503
Joined: 15-December 02
Member No.: 4082



Bit reservoir has nothing to do with frame sizes, which are always fixed to specific sizes. Bit reservoir means that a given frame may depend on data from previous frames.

Perhaps you would have an easier time understanding how this works if you look at the source code to a decoder instead of reading standards documents? I suggest looking at libmad if that is the case.
Go to the top of the page
+Quote Post
halb27
post Aug 30 2012, 09:04
Post #3





Group: Members
Posts: 2414
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



Bit reservoir is used with CBR (especially CBR 320), ABR, VBR.
Bit reservoir is the data space within a frame which is not used for audio data of the current frame. It can be used for audio data storage of the subsequent frame(s) - within specific limitations. This way bitrate can locally be roughly 50% higher than 320 kbps. It's all about frame data usage and is not restricted by CBR 320 or any other encoding mode.

As for choosing the "right" bitrate: The encoding machinery doesn't know the "right" bitrate, it just tries to do it right. The quality measurement of VBR and CBR is different, and for low to moderate quality levels VBR should be better in the overall view as long as the machinery doesn't have major flaws. For high to very high quality demands it's hard to say whether VBR or CBR is better overall. Opinion of most people probably is in favor of CBR 320 over -V0, but it was shown on HA already that -V0 can give better results than CBR 320.
And when it come to special solutions like my 3.99.5y -V0+ variant which maximizes bit reservoir, use extremely high audio data bitrate for impulses, and prevents audio data bitrate from going low (as CBR does) - and nonetheless is a VBR method -, things are becoming even more unclear as for which very high bitrate method is best in terms of quality.

This post has been edited by halb27: Aug 30 2012, 09:25


--------------------
lame3100m --bCVBR 300
Go to the top of the page
+Quote Post
krafty
post Aug 30 2012, 16:29
Post #4





Group: Members
Posts: 266
Joined: 20-March 10
Member No.: 79175



QUOTE
but it was shown on HA already that -V0 can give better results than CBR 320.


can you point me those links

So BR can happen on either CBR or VBR, but in VBR it is more relaxed...

I noticed also that there is no lowpass for V0 now, I wonder if this can lead up to some noise or is it actually an improvement over 320 CBR encodings...
Go to the top of the page
+Quote Post
halb27
post Aug 30 2012, 21:23
Post #5





Group: Members
Posts: 2414
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



a) a sample for which -V0 was preferred over CBR320

b) Bit reservoir usage isn't more relaxed with VBR. Instead, audio data bitrate variation of VBR relies more heavily on the variation of frame bitrate.

This post has been edited by halb27: Aug 30 2012, 21:24


--------------------
lame3100m --bCVBR 300
Go to the top of the page
+Quote Post
krafty
post Aug 31 2012, 20:36
Post #6





Group: Members
Posts: 266
Joined: 20-March 10
Member No.: 79175



So, as I understand... there were cases when V0 was better than 320. But there weren't any to show the other way around (320 better than V0) to at least guarantee a tie?
Go to the top of the page
+Quote Post
halb27
post Aug 31 2012, 22:39
Post #7





Group: Members
Posts: 2414
Joined: 9-October 05
From: Dormagen, Germany
Member No.: 25015



lead-voice is a sample that shows that CBR320 can be better than -V0.
Samples like these - together with the fact that -V0 as well as CBR320 usually yield an overkill quality - just show that it's hard to say what's best.


--------------------
lame3100m --bCVBR 300
Go to the top of the page
+Quote Post
krafty
post Aug 31 2012, 22:53
Post #8





Group: Members
Posts: 266
Joined: 20-March 10
Member No.: 79175



So theoretically we have a tie and should go for V0 because of less disk space.
Go to the top of the page
+Quote Post
greynol
post Sep 1 2012, 00:57
Post #9





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



As not to give the wrong impression...

"We" don't have a tie, one may be better than the other overall for any individual who can successfully demonstrate a difference. Furthermore, preference between two non-transparent encodings is absolutely subjective.

One should test for himself before making a conclusion as to which is better from a quality perspective and determining whether he belongs to the group you've designated as "we." Let's say that some particular individual cannot ABX anything including the most killer sample at -V2. Is -V0 better for him than -V1? The answer is no.

This post has been edited by greynol: Sep 1 2012, 01:17


--------------------
Your eyes cannot hear.
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: 20th April 2014 - 10:21