Beginners Guide To Ripping With Gordian Knot

MCMLXXXVIII

DxaKrator wrote:

Heheeheeheheeh
I get the same thing man..on both machines I encode on...not sure what bug it is or what...but I got to click my mbs down a bit on every rip otherwise it comes out oversized by about 5 to 8 mbs..depending on the size of the rip...No need to re-install or have a heartattack :lol:

Just tick it down to about 741 for 1/6 rips and about 1487 for 1/3..works like a charm for me.

Yes, I'd considered that. I'll try it next probably and see if I can get it released tonight.

This is the first time this has happened though, normally my AC3 rips come out bang on 1.45 GB.

Also, am I right in thinking that version 0.35.0 from Sep 2005 is the latest version? Strange that it has no official website and has never been updated since then. :roll:

video-man

reinstall xvid see if that helps

elchupacabra

I don't think there is a problem here. 8 MB is not much. If you set 1495 MB and subtract the audio size that is the target size XviD tries to reach in the end. My encodes are always off +/- 10 MB. I think that is normal, because that file size is just an approximation. The codec tries to do it's best to reach it and alternates the current bitrate accordingly during the encoding process.

It's like this: while you encode something how should you know the bitrate to use for the current frame if you only know how big the final video stream should be? All you can do is try to approximate. I mean there is a first pass to analyze the video but that's not perfect either.

elguaxo

MCMLXXXVIII wrote:

I've tried 3 times now to rip La Terza Madre and each time I get the same oversized (by 8 MB) rip.

How are the quants?

Open your oversized encode with AVInaptic then Statistics -> DRF analysis and post the results.

MCMLXXXVIII

I tried lowering the desired filesize by 8 MB... but it has ended up as 1.53 GB again.

Okay elguaxo, hope this makes sense to you:

[ About file ]

Name: madre.avi
Date: 11/05/2008 18:54:01
Size: 1,644,087,296 bytes (1567.924 MB)

[ Generic infos ]

Play duration: 01:37:12 (5832.48 s)
Container type: AVI OpenDML
Number of streams: 2
Type of stream nr. 0: video
Type of stream nr. 1: audio
Audio streams: 1
ISFT: VirtualDubMod 1.5.10.2 (build 2540/release)
JUNK: VirtualDubMod build 2540/release

[ Relevant data ]

Resolution: 704 x 288
Width: multiple of 32
Height: multiple of 32
Average DRF: 2.687160
Standard deviation: 1.006979
Std. dev. weighted mean: 0.597949

[ Video track ]

FourCC: xvid/XVID
Resolution: 704 x 288
Frame aspect ratio: 22:9 = 2.444444
Pixel aspect ratio: 1:1 = 1
Display aspect ratio: 22:9 = 2.444444
Framerate: 25 fps
Number of frames: 145812
Stream size: 1,310,433,350 bytes
Bitrate: 1797.428675 kbps
Qf: 0.354606
Key frames: 1320 (0; 300; 352; 355; 357; ... 145513)
Null frames: 0
Min key int: 1
Max key int: 300
Avg key int: 110.463636
Delay: 0 ms

[ Audio track ]

Audio tag: 0x2000 (AC3)
Bitrate (container): 448 kbps CBR
Channels (container): 6
Sample rate (container): 48000 Hz
Chunks: 145791
Stream size: 326,597,152 bytes
Preload: 448 ms
Max A/V diff: 472 ms
Initial skip: 1,344 bytes
AC-3 frames: 182252
Play duration: 01:37:12 (5832.064 s)
Chunk-aligned: No
Mode: 3 front, 2 rear, 1 LFE
Delay: 0 ms

[ About MPEG4 encoding ]

User data: DivX503b1393p
User data: XviD0046
Packed bitstream: Yes (*)
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: H.263
Number of frames: 145812
Drop/delay frames: 0
Corrupted frames: 0

I-VOPs:  1320 (  0.905 %)                          
P-VOPs: 53307 ( 36.558 %) #########                
B-VOPs: 91186 ( 62.536 %) ################         
S-VOPs:     0 (  0.000 %)                          
N-VOPs:     0 (  0.000 %)                          

Max consecutive B-VOPs: 2
    1 consec: 12436 ( 24.003 %) ######                   
    2 consec: 39375 ( 75.997 %) ###################      

[ DRF analysis ]

Average DRF: 2.687160
Standard deviation: 1.006979
Max DRF: 4

DRF=1: 19131 ( 13.120 %) ###                      
DRF=2: 46478 ( 31.875 %) ########                 
DRF=3: 41080 ( 28.173 %) #######                  
DRF=4: 39124 ( 26.832 %) #######                  
DRF>4:     0 (  0.000 %)                          

I-VOPs average DRF: 1.178787
I-VOPs std. deviation: 0.383174
I-VOPs max DRF: 2

P-VOPs average DRF: 1.661451
P-VOPs std. deviation: 0.473216
P-VOPs max DRF: 2

B-VOPs average DRF: 3.308621
B-VOPs std. deviation: 0.673976
B-VOPs max DRF: 4

[ Profile compliancy ]

Profile to check: MTK PAL 6000
Resolution: Ok
Framerate: Ok
Buffer underflow: 00:00:56 (frame 1411)
Buffer underflow: 00:00:57 (frame 1416)
Buffer underflow: 00:00:57 (frame 1421)
Buffer underflow: 00:00:57 (frame 1426)
Buffer underflow: 00:00:57 (frame 1432)
Buffer underflow: 00:00:58 (frame 1438)
Buffer underflow: 00:00:58 (frame 1447)
Buffer underflow: 00:00:58 (frame 1455)
Buffer underflow: 00:00:58 (frame 1462)
Buffer underflow: 00:00:59 (frame 1472)
Buffer underflow: 00:01:03 (frame 1587)
Buffer underflow: 00:01:04 (frame 1603)
Buffer underflow: 00:01:05 (frame 1627)
Buffer underflow: 00:01:06 (frame 1654)
Buffer underflow: 00:01:07 (frame 1680)
Buffer underflow: 00:01:08 (frame 1698)
Buffer underflow: 00:01:08 (frame 1712)
Buffer underflow: 00:01:09 (frame 1724)
Buffer underflow: 00:01:12 (frame 1811)
Buffer underflow: 00:01:25 (frame 2133)
Error: Too many violations

This report was created by AVInaptic (18-11-2007) on 11 mag 2008, h 19:02:26

elguaxo

MCMLXXXVIII wrote:

Okay elguaxo, hope this makes sense to you:

Yes it does!

1) Quant type: H.263

H.263 is evil, it tends to block even if the compressibility is great. Use MPEG or a CQM. Try also BVOPs at 2/1.62/0

2) DRF=1: 19131 ( 13.120 %) ###

this is VERY bad. Quants=1 just fill size but won't give you more quality or details. Use them at min=2

Try again and report back! :beerchug:

MCMLXXXVIII

Thanks elguaxo, I'll change the settings you've indicated and try it again tonight. :wink:

MCMLXXXVIII

It's undersized now: 1.12 GB.

video-man

u having trouble on a pal dvd

MCMLXXXVIII

video-man wrote:

u having trouble on a pal dvd

Yes it is. How might this matter?

elguaxo

MCMLXXXVIII wrote:

It's undersized now: 1.12 GB.

That was expected! :wink:

Post the complete DRF analysis of your new encode. The settings I suggested aim at reaching the sweet spot of most CQMs or MPEG in this case.

video-man

MCMLXXXVIII wrote:

video-man wrote:

u having trouble on a pal dvd

Yes it is. How might this matter?

nothing but i have had problems to with pal dvds being oversized on the encode but when i do a ntsc it turns out a perfect 1.46

MCMLXXXVIII

Okay, here goes:

[ About file ]

Name: madre.avi
Date: 12/05/2008 01:00:46
Size: 1,203,714,048 bytes (1147.951 MB)

[ Generic infos ]

Play duration: 01:37:12 (5832.48 s)
Container type: AVI OpenDML
Number of streams: 2
Type of stream nr. 0: video
Type of stream nr. 1: audio
Audio streams: 1
ISFT: VirtualDubMod 1.5.10.2 (build 2540/release)
JUNK: VirtualDubMod build 2540/release

[ Relevant data ]

Resolution: 704 x 288
Width: multiple of 32
Height: multiple of 32
Average DRF: 2.621144
Standard deviation: 0.485101
Std. dev. weighted mean: 0

[ Video track ]

FourCC: xvid/XVID
Resolution: 704 x 288
Frame aspect ratio: 22:9 = 2.444444
Pixel aspect ratio: 1:1 = 1
Display aspect ratio: 22:9 = 2.444444
Framerate: 25 fps
Number of frames: 145812
Stream size: 870,060,194 bytes
Bitrate: 1193.399986 kbps
Qf: 0.235440
Key frames: 1326 (0; 300; 352; 355; 357; ... 145513)
Null frames: 0
Min key int: 1
Max key int: 300
Avg key int: 109.963800
Delay: 0 ms

[ Audio track ]

Audio tag: 0x2000 (AC3)
Bitrate (container): 448 kbps CBR
Channels (container): 6
Sample rate (container): 48000 Hz
Chunks: 145791
Stream size: 326,597,152 bytes
Preload: 448 ms
Max A/V diff: 472 ms
Initial skip: 1,344 bytes
AC-3 frames: 182252
Play duration: 01:37:12 (5832.064 s)
Chunk-aligned: No
Mode: 3 front, 2 rear, 1 LFE
Delay: 0 ms

[ About MPEG4 encoding ]

User data: DivX503b1393p
User data: XviD0046
Packed bitstream: Yes (*)
QPel: No
GMC: No
Interlaced: No
Aspect ratio: Square pixels
Quant type: MPEG
Number of frames: 145812
Drop/delay frames: 0
Corrupted frames: 0

I-VOPs:  1326 (  0.909 %)                          
P-VOPs: 53916 ( 36.976 %) #########                
B-VOPs: 90571 ( 62.114 %) ################         
S-VOPs:     0 (  0.000 %)                          
N-VOPs:     0 (  0.000 %)                          

Max consecutive B-VOPs: 2
    1 consec: 14021 ( 26.811 %) #######                  
    2 consec: 38275 ( 73.189 %) ##################       

[ DRF analysis ]

Average DRF: 2.621144
Standard deviation: 0.485101
Max DRF: 3

DRF<2:     0 (  0.000 %)                          
DRF=2: 55242 ( 37.886 %) #########                
DRF=3: 90571 ( 62.114 %) ################         
DRF>3:     0 (  0.000 %)                          

I-VOPs average DRF: 2
I-VOPs std. deviation: 0
I-VOPs max DRF: 2

P-VOPs average DRF: 2
P-VOPs std. deviation: 0
P-VOPs max DRF: 2

B-VOPs average DRF: 3
B-VOPs std. deviation: 0
B-VOPs max DRF: 3

[ Profile compliancy ]

Profile to check: MTK PAL 6000
Resolution: Ok
Framerate: Ok
Buffer underflow: 00:00:57 (frame 1419)
Buffer underflow: 00:00:57 (frame 1432)
Buffer underflow: 00:00:58 (frame 1454)
Buffer underflow: 00:22:04 (frame 33099)

This report was created by AVInaptic (18-11-2007) on 13 mag 2008, h 18:16:34

elguaxo

Ok! With BVOPs at 2/1.62/0 Xvid tends to saturate at DRF ~2.5, so your undersized encode is normal. Even if the size is smaller the quality is better than your first oversized encode because:

Before:
Quant type: H.263 -> tends to blurr and block

Now:
Quant type: MPEG -> keeps more details

Before:
Standard deviation: 1.006979
Std. dev. weighted mean: 0.597949

Now:
Standard deviation: 0.485101
Std. dev. weighted mean: 0
which means your current encode is more 'stable', for example no 'swimming walls'.

And now you want to nail it at 1.45 GB :mrgreen:

That's easy, keep your current settings, but instead of Quantization type: MPEG use this CQM: EQM V3HR. The quants will be higher than your current MPEG version, but this matrix keeps even more details.

Edit: there is still a chance that you will get an undersize. Use EQM V3HR as suggested, but set the BVOPs at 1/1.62/0:

I hope that's enough, if not, then try HEINI_MR instead of EQM V3HR.

MCMLXXXVIII

Okay I think I understand that! One thing, where do xcm files go?

I'll try your suggestions, hopefully I'll get the right size and a better rip.

elguaxo

Here and then Load matrix:

All the settings in this window are for the 1st and 2nd pass.

spudthedestroyer

elguaxo wrote:

Ok! With BVOPs at 2/1.62/0 Xvid tends to saturate at DRF ~2.5, so your undersized encode is normal. Even if the size is smaller the quality is better than your first oversized encode because:

Before:
Quant type: H.263 -> tends to blurr and block

Now:
Quant type: MPEG -> keeps more details

Before:
Standard deviation: 1.006979
Std. dev. weighted mean: 0.597949

Now:
Standard deviation: 0.485101
Std. dev. weighted mean: 0
which means your current encode is more 'stable', for example no 'swimming walls'.

And now you want to nail it at 1.45 GB :mrgreen:

That's easy, keep your current settings, but instead of Quantization type: MPEG use this CQM: EQM V3HR. The quants will be higher than your current MPEG version, but this matrix keeps even more details.

Edit: there is still a chance that you will get an undersize. Use EQM V3HR as suggested, but set the BVOPs at 1/1.62/0:

I hope that's enough, if not, then try HEINI_MR instead of EQM V3HR.

Can I ask you to show your maths please, how did you work out the bvops ?

Quote:

With BVOPs at 2/1.62/0 Xvid tends to saturate at DRF ~2.5

What people would like to know is the relationship between the BVOP settings and the DRF, and a ranking of matrices. I've been unable to find one myself.

You only get 1 mark for the answer, and 5 for the working out. :googley:

elguaxo

spudthedestroyer wrote:

Can I ask you to show your maths please, how did you work out the bvops ?

The math is beautiful

It gives you Bquant = Pquant + 1 for small quantizers (2 & 3, see for example the very tight quant distribution in MCMLXXXVIII's undersized sample using this setting), and for higher quants (q5 up to q10) is almost identical to XviD's defaults (2/1.5/1). And 1.62 is aprox. the legendary "golden cut". :mrgreen:

Code:

          2,0 + 2,0
BQuant = ----------- x 1.62 + 0 = 3.24 = which gives you a BVOP of 3
              2
    

          2,0 + 3,0
BQuant = ----------- x 1.62 + 0 = 4.05 = which gives you a BVOP of 4
              2

and for higher quants it's almost the same as the default

          4,0 + 5,0
BQuant = ----------- x 1.5 + 1 = 7.75 = which gives you a BVOP of 7
              2


          4,0 + 5,0
BQuant = ----------- x 1.62 + 0 = 7.29 = which gives you a BVOP of 7
              2

spudthedestroyer wrote:

What people would like to know is the relationship between the BVOP settings and the DRF, and a ranking of matrices. I've been unable to find one myself.

I'm just interested in doing quality encodes and they have most of the quants at 2/3/4 if you do the first pass at Q2 or 3/4 and some 5s if you do the first pass at Q3 and use a matrix like 6of9 or above. Both scenarios are reached with 2/1.62/0 (or 1/1.62/0 if you can afford it). With the first pass at Q2, which is what we use almost all the time, the lower limit is ~DRF 2.5. Why? All I-Frames will be 2, all P-Frames will be 2 and all B-Frames will be 3, B-Frames will be aprox %50 to %60 of all frames and you get this result. Again, see MCMLXXXVIII's undersized sample above.

Matrices: there are many nice matrices, but to simplify and ordered by Less Details/more compressibility to More Details/Less Compressibility:

Less Details/More Compressibility
Jawor's 1CD (or Soulhunters v8)
EQM V3LR
Jawor's ZSM
EQM V3HR
Heini_MR
6of9
More Details/Less Compressibility

If you get an undersize with one, use the next bigger one. If you get an oversize, use the next smaller one.

Recommended reading:
http://forum.doom9.org/showthread.php?t=119399
http://forum.doom9.org/showthread.php?t=93539
http://forum.doom9.org/showthread.php?t=119500

elchupacabra

Oh boy thanks a lot mate. This sounds very interesting. I'll read these when I find some time. :beerchug:

spudthedestroyer

perfect, thank you elguaxo :beerchug:

Beginners Guide To Ripping With Gordian Knot

Who is online