Interesting failure in CCE 2.50

[Kedirekin]

Over the past few days, I've been trying to encode Urusei Yatsura #2: Beautiful Dreamer in CCE 2.50 without success. I thought I'd share the strange problem I'm experiencing; CCE simply exits part way into the first (vaf) pass. It doesn't generate any error messages, and it doesn't simply hang. It closes/exits as if I had hit cancle and selected File/Exit. It leaves behind a partial vaf file, but otherwise it's as if I'd never run it in the first place.

At first I thought it might be AviSynth that was causing the problem, so I tried a number of things to work around it. I tried:
- d2v->avs->CCE (original approach, of course)
- d2v->avs->vDub->CCE
- d2v->avs->vDub->avi (oddly, this failed while saving the avi*)
- vDub->avi, then avi->avs->CCE
- vDub->avi, then avi->avs->vDub->CCE
and finally,
- vDub->avi, then avi->vDub->CCE (abandoning Decimate and encoding at 29.97 fps)
(I didn't try any VFAPI approaches - if it doesn't work via vDub, I see no reason to expect it to work via VFAPI)

All of these approaches failed at about the same point, approximately 81 minutes into the movie (based on the frame count in the vDub frameserver dialog and the relative sizes of the partial vaf files). The last one finally convinced me it wasn't an AviSynth problem (AviSynth isn't even in the mix) - it truly had to be something in CCE.

I looked at the video at 81 minutes and found a yellow strobing effect that switches on and off every two frames. This effect goes on for about 20 seconds.

I surmised (in other words, guessed) that CCE was freaking out because it's recognizing a scene change every two frames in this section of footage, and simply can't handle it (either because of the low bitrate (300/1950/2400) or because it can't handle a succession two-frame GOPs). In desparation, I turned on 'Disable auto I-frame insertion' (because as I understand it this shuts off scene change detection) and tried encoding again. Viola, success (though I can't speak for the quality of the encode - I'm burning it now).

Well, that's my story. Has anyone else exprerienced this sort of problem? Did/do you have a different guess as to what the real problem is? Any suggestions on a better work-around? Do you think a later version of CCE might handle it better?

Any reply appreciated. TIA.

----------------------
* I think this strange avi save failure is related to a strange avs->vDub 2BG problem. It isn't a file system problem (I'm using NTFS), nor is it an avi file limitation (it fails with multi-segment as well). I simply can't process avs-to-avi via vDub if the end file(s) total larger than 2 BG. I have no idea why.

[Kedirekin]

Just thought I'd mention that this just happened to me again; CCE simply exited half way through the vaf pass with no error message. This time it happened on episode 12 of Nazca.

I immediately tried encoding again with 'disable auto I-frame insertion' checked, and the encode succeeded.

I ran vDub as a frame server and found that CCE failed at frame 18968 (from the start of the first chapter of episode 3, at 23.97 fps). Looking at the video near that frame, I found that it was undergoing rapid scene changes. From frame 18783 to 19021 there is a complete scene change every 7 or 8 frames (238 frames, 9.92 seconds, about 30 scene changes). Before that, there is this strange footage of an attacking character's face bouncing erractically around the screen about every 3 frames - erractically enough I'm sure to cause scene change detection (this starts at frame 18594, for a total of 472 frames, 19.7 seconds, bumping scene changes to about 93).

This is not like the strobing I described in Urusei Yatsura #2; these are real scene changes (it's a flash-back (or flash forward, depending on how you look at time and space) remembering individual moments of a persons life, plus the flashing face).

Now the fact that this has happened to me twice in such rapid succession (and it has never happened before) makes me think there's something wrong with my PC (despite the fact that I've done other encodes between and since with no problem).

On the other hand, the fact that the failure happens in such similar circumstances (footage that would cause CCE to insert a lot of small GOPs) makes me think this is a flaw in CCE. And the fact that shutting off 'auto I-frame insertion' allows the encode to succeed strengthens this hand. Perhaps the bitrate averaging algorythms of the vaf pass freak out when there are not enough P and B frames to keep the bitrate below the max.

So, I wanted to ask:
- has anyone ever had CCE simply shut down with no error message and for no apparent reason?
- can anyone confirm they've successfully encoded Urusei Yatsura #2 or Nazca episode 12 normally with CCE without problem?

BTW: checking the 'disable auto I-frame insertion' checkbox in CCE has minimal (if any) impact on the quality of the encode. I know that's counter-intuitive, and obviously I haven't been able to do a side-by-side comparison with these titles, but compared to my nominal encodes there is little or no difference.