View Full Version : TMPGEnc setting question

07-08-2005, 11:52 AM
Hi Y'all (especially Eugene).

I've been running into a rare TMPGEnc problem with a clip my client generated. It's an animation that is rendered in very dark shades of blue. At the end of the clip it fades to black over 3 seconds. In the source RTV, it fades out very nicely. When I encode it with TMPGEnc, the final fade "strobes" very noticeably, like it doesn't have enough colorspace so it quantizes the blues on the way to black, OR it's not picking up the subtle shading differences between the i-frames.

My question is are there any settings in the TMPGEnc 2.5 panels that may help here? I'm using Eugene's standard settings at 6000VBR. I've tried the "CG/animation" matrix but it does the same thing. Are there any other settings I should try?

Thanks! -MG.

07-08-2005, 12:35 PM
You may want to consider using CBR instead of VBR. You never know sometimes where TMPGEnc thinks it can drop down to a lower data rate when it shouldn't. CBR forces TMPGEnc to maintain the data rate you've chosen. I'm not sure this will solve your issue but it may be worth a try.

07-08-2005, 01:01 PM
You might want to try manually setting more I-frames in that dissolve. TMPGEnc lets you do that, it might help.

Of course, being that the MPEG-2 is 4:2:0 colorspace versus 4:2:2, that could be part of the problem.

Though honestly, every cause of strobing in MPEG-2 that I've seen has been due to the wrong fielding.

07-08-2005, 01:49 PM
Heh. "Strobing" probably isn't the best word for it. The background of this animation is jumpback-type swirling patches of color, and instead of fading out smoothly in the MPEG, they stairstep down to black. The various colors stairstep at different points which is what makes it strobey. The audience probably wouldn't notice, but the client definitely would. (Let's see, which new smiley... this one! :censored: ).

Thanks very much for the suggestions, I'll try them out right now. I think Eugene's on the right track that I could be asking too much of the colorspace (I forget, is the "0" in 4:2:0 = blue, or V?)...

Thanks! -MG. :beerchug:

07-08-2005, 03:05 PM
Actually, neither. If the "0" was blue or red, as in those colors aren't there, we'd have a big problem! ;)

4:2:2 and 4:1:1 are more linearly related when you draw the actual matrix of what's happening. 4:2:0 you need to actually see the matrix and be thinking two dimensionally instead of one dimensionally for it to make sense. I don't really know how to describe it better, but if you do a web search, it's an interesting read.

07-08-2005, 06:00 PM
When you put it that way, I guess we would have a problem. :eek: (We could always get out the red/green 3D glasses...)

Thanks for the brief explanation. I just tried a bunch of encodes including CBR at high bit rates and adding lots of I-frames (every other frame) to the area, but the problem persists. It doesn't seem to be a lack of bits, I think I've just walked off the edge of the gamut.

Thank you gentlemen for your help, as always! -MG. :thumbsup: