KDVD: Best CCE matrix settings?
 

To encoder a film with 4000~5500 Kbps of scale is better use matrix with less compression even enlarge the factor Q in CCE? Or the inverse?
Sorry, my english is bad...

Use the kvcd-Notch matrix and set the Q factor as needed to fill your disc.

KVCD_Notch
 

Yes, Notch is a good Matrix, but my hesitation is: to high bitrates CCE Standart Matrix ( or MPEG) is not better? >>>1 film to DVD-5

I understood the question and I answered yet : use the Notch ;)
MPEG standard matrix is far worse, I don't remember to have fully tested the CCE one.

Thank you!

Notch is optimized matrix for TMPGEnc but certainely not for all MPEG2 encoder. TMPGEnc in CQ mode use very constant macroblock quantizer (for I,P and B and for inter/intra block) but not CCE. Notch done very good result with TMPGEnc in CQ mode but not for multipass mode (in theory the best mode for quality repartition).

Noth matrix is really good for blocking but not for ringing : For my eyes, Notch is really too agressive for high frequencies and produce blur even with low quantizer. For low quantizer (aka high bitrate) is not very good Custom Matrix.

Actually, if you set the high inter/intra quadrant to the same values of the standard MPEG matrix (if you divide the matrix in 4 equal pieces, and take the bottom right ), then your high frequency domain will be identical to standard matrix, while still maintainning the notch advantage on low lit/dim conditions ;)

-kwag

My first attempt with HC encoder prooved that this theory is nothing more. Let the theory where it has to be : far from practical reality ;-).

I can agree with that. Nevertheless, compared to CCE default matrix or even the orignal MPEG matrix, it is far better. But there are probably better matrixes, depending on the encoder as you mentioned.

I make visual test with CCE and HCEnc and notch work not really good in multipass mode : these encoder use higher ratio/offset for bframe and inter matrix for Notch is really too agressive for that (high relative average coef + higher quantizer -> done high relative ringing level and poor relative quality in low motion part).

In fact inter/intra matrix coef and inter/intra macroblock setting (AQ, ration/offset for I,P,Bframe, rate control curve compression ...) aren't indepandant setting. TMPGEnc, Libavcodec, HCEnc, CCE and other don't use same internal setting for inter/intra macroblock and thus CQM matrix will not done the same final result for all MPEG2 encoder.

All I wanted to say is that multipass mode is not the best mode (at least for these two encoders - and for tmpgenc also), whatever the theory.

And I second that ;)

-kwag