4 passes m2v a lot bigger than first pass m2v?

[StephLG]

Hi,

I've just encoded a 126 minutes movie (NTSC, 720x480@23.976fps from Laserdisc capture) with CCE v2.62 and there's something that strikes me; the resulting m2v file from the 3 passes process (4 if we count the first pass) is a lot bigger than the m2v file from the first pass: 3.8GB vs 1.2GB :eek:
I basicaly followed Robshot settings (in his 'Getting the best out of CCE' guide) except for the 'linear quantization scale' that I unchecked because it is not meant to be used for Mpeg2 (at least, that's what they say in the manual) and it sometimes produced some overflow errors during the second pass.

So, the setting were:
1st pass VBR: Q=60 min=500 max=8000 no audio
3 passes VBR: min=500 avg=4500 max=8000 no audio

I didn't make any bitrate adjustment between first pass and the next 3 passes because the graph (in advanced tab) looked normal with no grey area.
I thought the next 3 passes would allocate more efficiently the bits but would keep approximatly the same overall size. Does this result (3.8GB vs 1.2GB) mean that the first pass was enought and that I just 'wasted' bits by setting the average bitrate too high? Or does it mean that the first pass under-estimated the bits needed because of a Q factor too hight?

I didn't watch the m2v from the first pass (maybe I should have :( ).

Can someone give me some advice?

Thanks,
Steph.

[Pyre8]

What was the average bitrate from the first pass? If CCE did the first pass at a lower average bit rate (say 2000) and then you raised the average bit rate to 4500 then the resulting file would get bigger. Just a thought.

If you did not make any bit rate adjustments between the first pass and the next passes, then you could have done multi-pass VBR without doing a first pass (since you are letting CCE determine bit rates based on a Q, min and max). Robshot's method is best when you want to lower the average bit-rate from what CCE determines during the first pass and then 'adjust' bit rates in only the 'grey' areas.

[StephLG]

Hi Pyre8,

I'm not sure I understand what you say; there's no average bitrate setting for the first pass. Only Q factor, min and max bitrate.
What I did is:
- one pass VBR to obtain a .vaf file (and a 1.2GB m2v file) with Q=60, min=500 and max=8000
- multi-pass VBR (with passes set to 3) to obtain the final 3.8GB m2v with min=500, avg=4500 and max=8000.

[Pyre8]

What was the average bit rate at the end of the first pass? Did you enter/change the average bit rate to 4500 for the multi-pass?

[StephLG]

Originally posted by Pyre8
What was the average bit rate at the end of the first pass?
That's what I don't understand. Where is this average bitrate available? In the advanced window when you switch to multi-pass VBR? If yes, I didn't look at it. :confused:
Did you enter/change the average bit rate to 4500 for the multi-pass?
Yes, I entered 4500 manualy for the multi-pass, based on nothing but the feeling that it would be enough but not too much (actualy, it's a little bit too much since I'd like to have a 3GB file instead of 3.8GB).

I think I'd better go again through the whole process and pay more attention to the details ;)

Thanks for your help.

Steph.

[auenf]

open up the first pass in bitrate viewer and you will find that it is at a fair bit lower than your 4.5mbit, what you should do is start at 2mbit to do the robshot method properly.

Enf...

[Pyre8]

At the end of the first pass, when you click on the Multi-pass button, the options change from the Q/min/max to average/min/max (just below the Multi-pass option box). I thought that the average shown at that stage was the average that CCE had used in its first pass. Or is it just a starting value that CCE defaults to when doing a multi-pass?

Okay, at least we know why the file size grew now;)

[mb1]

bitrate viewer. I only use it to see the quantization distribution and say "Oh! CCE is an awesome encoder."
Then Ligos LSX would be an outstanding encoder (which it isn't) ...

In fact using comparable settings lsx always produces lower (= better) quantisation values than cce (same source material of course).

[chainsaw135]

yea i've used the (ligos lsx) a few times and i really didn't like it. No matter the picture was blocky and the time it took to finish the movie was slow.