Settings for 1-click mode in DVD-RB

[Potlood]

I want to back-up some of my DVD9s to DVD5 and I have some questions about quality vs. the ease of 1-click mode. What settings are 'safe' for most movies so I can apply them once and then just forget about the whole settings thing?...

I did some testing (retail dvd: Blood Diamond) and these are the settings I used with CCE:

BACKUP MODE: Movie only
BITRATE REDISTRIBUTION: Enabled
PASSES: 3 (1 .vaf + 2 VBR)
VBR BIAS: 0 (http://forum.doom9.org/showthread.php?t=128884)
QUAL PREC: 16
MATRIX: Extract from movie and use CCEAQM=1
FILTERS: UnDot()
OTHER SETTINGS: Default


what would give beater quality in theory?

1 - Using standard matrix with CCEAQM=1
2 - Extract matrix from movie
3 - Extract matrix from movie and use with CCEAQM=1
4 - Use some custom matrix (not tried)

I did some testing and in my opinion nr3 looked best. But that's just one man's opinion..
If extracting a matrix really does give better quality in all re-encodes, is there a way to do it automatically?

Notice I set VBR BIAS to 0. Does it affect CCEAQM=1? (Maybe more changing in matrix) And if it does, is that a good thing?

I also read something about cpu=4 parameter in mpeg2source("................",cpu=?) to reduce macroblocking. Could setting that improve

quality even more?



Thanx

[Potlood]

Nobody? :confused:

[chickenmonger]

My guess is that the best way to really determine would be to do some testing, and post some screenshots for the community. Of course, that's not really an answer to your question, but it does reinforce the subjective matter of DVD backup.

As for the mpeg2decode option, the manual (http://neuron2.net/dgmpgdec/DGDecodeManual.html#MPEG2Source) says:

cpu: 0 to 6 (default: 0)

Post-Processing Quickset Options.
(Y=luma, C=chroma, H=horizontal, V=vertical)
- 0: DISABLE POST-PROCESSING
- 1: DEBLOCK_Y_H
- 2: DEBLOCK_Y_H, DEBLOCK_Y_V
- 3: DEBLOCK_Y_H, DEBLOCK_Y_V, DEBLOCK_C_H
- 4: DEBLOCK_Y_H, DEBLOCK_Y_V, DEBLOCK_C_H, DEBLOCK_C_V
- 5: DEBLOCK_Y_H, DEBLOCK_Y_V, DEBLOCK_C_H, DEBLOCK_C_V, DERING_Y
- 6: DEBLOCK_Y_H, DEBLOCK_Y_V, DEBLOCK_C_H, DEBLOCK_C_V, DERING_Y, DERING_C

So it just enables some built in deblocking. Your DVD sources would probably be fairly clean as far as blocking goes. If the source was blocky, however, it might be handled better with an AviSynth plugin instead.

[Sharc]

I agree that you have to make your own experience and see what makes the difference for you. There is no absolute rule or truth here as much is subjective and moreover depends on the source material. Leaving everything at the default settings produces normally very decent results for DVD backups.
Just a word of caution:
- I would be reluctant to just adding any filters when you have no specific reason to do so. Even a simple filter like undot() can produce undesired effects on certain scenes.
- Redistribution is a good thing and helps to get most out of the backup, however you risk to run into an CCE stability problem when used with high bitrate matrices. The result is that some segments may become blocky. You find more about this in the related threads on redistribution and the discussion about the CCE crash there. Extracted matrices are often of the high bitrate type (eg FHE), hence to be on the safer side I would use the standard matrix for redistribution, or switch to HC.

Just my two cents.

[Potlood]

I agree that you have to make your own experience and see what makes the difference for you. There is no absolute rule or truth here as much is subjective and moreover depends on the source material. Leaving everything at the default settings produces normally very decent results for DVD backups.

I did some more testing. Compared some scenes frame by frame using virtualdub (fast action, motion in dark scenes, explosion scenes, jungle scenes)

Got these results:

HCE with extracted matrix and LUMGAIN = 2 >> most macroblocks (sometimes really blocky compared to CCE)
HCE with default matrix and LUMGAIN = 2 >> not tested yet
CCE with default matrix and CCEAQM = 1 >> few macroblocks
CCE with extracted matrix and CCEAQM = 1 >> fewest macroblock

Found CCE producing slightly sharper images while less blocky then HCE
Quality was great with both encoders and those macroblocks were almost invisible while normal movie watching

- I would be reluctant to just adding any filters when you have no specific reason to do so. Even a simple filter like undot() can produce undesired effects on certain scenes.

Ok. I thought UnDot() could be used to increase compressibility without making major changes to the source

- Redistribution is a good thing and helps to get most out of the backup, however you risk to run into an CCE stability problem when used with high bitrate matrices. The result is that some segments may become blocky. You find more about this in the related threads on redistribution and the discussion about the CCE crash there. Extracted matrices are often of the high bitrate type (eg FHE), hence to be on the safer side I would use the standard matrix for redistribution, or switch to HC.

In my tests CCE standard matrix had more blocks then the extracted one (almost same as FHE). Perhaps adaptive matrix switching prevents that CCE stability problem?

BTW After rebuilding I once again extracted matrices. CCE matrix filesize was about 100 kb while HCE's was over 1.2 MB... How is that possible?

[steptoe]

I use the matrices AVAMAT6 and AVAMAT7 depending what the bitrate comes out as after the intial prepare phase as suggestions on other forums say these are good all round matrices for most sources

Also, try using removegrain(mode=1) instead of undot() as its faster and does exactly the same job and apparently undot() has a bug that removegrain(mode=1) doesn't have, but you'll have to install the removegrain filter to use it, plus you can choose faster versions depending on your CPU to speeed it up again

It also means if you have a multi-core CPU you can also use MT with removegrain to speed it up even more


Also, if you have plenty of memory, as suggested on the avisynth forums, set the first line in your filter editor to SetMemoryMax(128) if you have more than 1GB of memory, I use SetMemoryMax(512) with no problems but I have 2GB at present

Install avisynth 2.5.7, some filters and functions will NOT work without this version installed and there are quite a few bugs and memory issues fixed, it appears any problems there was have been worked out as they are now on 2.5.8 beta testing and jdobbs is considering upgrading the install to 2.5.7 anyway as it now appears to be a lot more stable than it used to be

[tom942]

- Redistribution is a good thing and helps to get most out of the backup, however you risk to run into an CCE stability problem when used with high bitrate matrices. The result is that some segments may become blocky. You find more about this in the related threads on redistribution and the discussion about the CCE crash there. Extracted matrices are often of the high bitrate type (eg FHE), hence to be on the safer side I would use the standard matrix for redistribution, or switch to HC.

Sharc, about redistribution, do you refer to the whole process (prepare, encode and rebuild) or just do the prepare phase with the standard matrix and the switch to other matrix upon the bitrate obtained?. (We are always referring to CCE).

[Sharc]

I refer primarily to the redistribution phase. After the redistribution phase you may switch to a matrix of your choice for the subsequent (2-pass) encode. You can select a new matrix with RB-Opt for example.

[jdobbs]

The next DVD-RB version includes an option to always use HC as the encoder for REDISTRIBUTION (only). That way you don't have to worry about the CCE bug and what matrix is selected -- but CCE will be used for encoding (when chosen).

[Sharc]

That will certainly cure some headaches ....

Btw: The matrix that is used for redistribution will be the one defined as "Main Feature Matrix" for CCE SP 2.70 and CCE SP2? From another post I remember that SP2 would always take the CCE default matrix, whereas for CCE 2.70 the "Main Feature Matrix" would be used? Not sure, though.

[tom942]

OK, I'll try what you said Sharc.

Until now, I've been doing both phases with the same matrix, for instance FHE matrix or other different than encoder's default.

Edit: At the moment, it is not possible to do the same thing with HC, is it?.

I mean, redistribution phase with HC default matrix and later switch to other matrix for encoding phase.

[Sharc]

With DVD-RB you have - at the moment - to make the decision at the beginning if you want to do everything - redistribution and encoding - with CCE or HC (with CCE Basic and Procoder being the exceptions).
I would wait for the next version of DVD-RB as announced by jdobbs in his recent post. As I understand it will allow you to use HC for the redistribution phase and CCE (or another encoder) for the subsequent multipass encoding. Matrix selection for the encoding phase will be according to your choice in Settings=>Options=>Advanced Options of DVD-RB Pro.
Using HC exclusively for the redistribution phase is actually a workaround for the annoying CCE bug.

[tom942]

In the new version (1.26.3), it still uses the matrix defined in "Main Feature Matrix".

Is there a way with HC (or other encoder) to use in the redistribution phase the "default matrix" and later, in the encodig phase, the one that I think it is most suitable?.

P.S.: I don´t say anything about CCE because I know I can change it with RB-OPT :).

[jdobbs]

If you use a different matrix than the one selected during REDISTRIBUTION it would seem the results of the REDISTRIBUTION wouldn't be valid???

[tom942]

If I don´t understand you wrong, are you telling me to use the same matrix in both phases?.

If yes, then I understood Sharc in the wrong way, thinking that I've got to use "encoder default" matrix for redistribution and then use a most suitable matrix for encoding.

Re-reading the posts, I think that Sharc was referring to the way to avoid CCE bug, no?

[jdobbs]

As Sharc mentioned, the better way to get around that bug with DVD-RB v1.26.3 would be to use HC for the REDISTRIBUTION and CCE for encoding when you are using risky matrices.

[tom942]

Then sorry, I understood it in the wrong way :(.