Help Needed: CCE generating invalid GOP structure

[SaSi]

The issue about GOPs and headers has been discussed a lot, but this is a different problem.

After encoding a 2 hour stream in CCE (v.2.62) with parameters correct for DVD and DVD compatibility checked, Scenarist refuses to import the stream complaining about a wrong GOP length (30) in GOP 4056. To my understanding, this meant that GOP 4056 was produced out of spec.

To verify this, I cut the first 50.000 frames (GOP 4056 should be near 60.000) and Scenarist accepted the first segment, meaning that the "general" stream structure (at least in the beginning) was correct.

Any idea why this is happening?

[damnim]

I'm guessing

First what are you setting your GOP sequences to be?
If it is above 18 "3x6" Scenarist won't take it and irritate you about it. "3x4" and "3x5" are really good settings and matches most commercial encodes.

What are you setting the packet size as?
You should set it to 2048 and if you want to use protection do that in Scenarist.

[SaSi]

The GOP settings in CCE are correct for DVD (M=3 and N/M = 5) and GOP Headers every 1 x N frames while SEQ headers every 1 GOP.

Since my posting I did some "research" of my own finding that:

On single pass CBR, no problem.
On single pass VBR, a random point deep down in the movie stream has invalid count of frames (30).
On 2 pass VBR, no problem.

The above was repeated 3 times each (quite a few hours of encoding) and although as a test it is not extensive, it appears that single pass VBR produces a randomly positioned invalid frame sequence two times out of 3.

In addition to Scenarist finding the error, DVD Fusion rejects the stream after scanning midway through the stream.

Have you seen such a problem yourself?

[damnim]

I've encoded more hours of video then I can count and I have run into this situation once before where Scenarist rejected the stream due to an invalid frame count within the GOP. That's why I asked about the GOP settings first. At the time I was unfamiliar with IBP frames within a sequence and had various frame counts. Once I corrected the settings I never had the problem again. But I wonder now;

Are you frameserving?
Are you using pulldown?
Is your entire program stream consistent ie;interlaced, progressive, top field, bottom field?
Are you restricting I frame insertion or closing the GOP's?
Are you multitasking while encoding?

I frameserve using avisynth and create a project with DVD2avi and have had no problems using this setup.
I don't use pulldown on original encodes because of the failure rate I was getting. CCE was doing some weird things to the frames and after setting the flags the picture was choppy.

I would try cutting the problem segment out of the elementary stream and encoding it with tmpgenc. If Tmpgenc encodes it correctly then my bet is that your CCE is corrupt or needs to be patched or updated.

[SaSi]

You got that @dannim!

The answer to all your suggestions is "YES" settings correct. However, it is correct to the question "Are you multitasking?"

I realize that multitasking messes up the whole thing in many applications. I have seen smartripper verbally rip a stream if the CPU is pressed too hard. Also seen the same thing in Tmpgenc.

Since my original post, I pulled back a little and have done 2-3 encodings in CCE. One was 2 pass VBR and the other 2 were single pass VBR. All were good. Oly difference was that I didn't touch the PC while encoding.

I will keep the issue open in my mind for a while but avoid stretching the PC while encoding. If this never re-appears, then all is well.

My problem was how much can I trust the quality of the transcoding, especially if it is a 4-5 hour process.

By the way, I am not frameserving. I have tried that a couple of times, just to understand it, and it appears to be a bit messy and unreliable.

Also, doesn't fit within the set of rules I am trying to configure for a professional studio. These people are used to put tapes into decks and press buttons. They will jump to the moon if I try to explain frameserving.

[damnim]

I went and looked at your hardware profile.

Check to see if your page file settings in Win 2000 is set to 1.5 that of your ram.

Also do you defrag regularly?

These things can cause significant problems.

I use a dedicated computor for encodes. FYI

[SaSi]

I have virtual memory configured at 1534 MB (~1.5 x RAM size) and I also have a couple of 20Gb partitions that are used for encodes. Encoding is done from a 200Gb disk to a 20Gb partition on a 40Gb disk.

Source avi is huffyuf format and ranges in size between 30Gb to 120Gb depending on avi length and content.

I have seen fragmentation kill performance quite a few times and it has been a while my systems are configured to defend agains that. I have a dedicated system volume (27Gb) that only holds applications, a Data Volume that has the swap file in the beginning and any data files (documents, etc) plus a few more disks. 2 x 70G and 2 x 200Gb.

FYI, I have done a couple more encode sessions with CCE since my last posting and both were correct streams. Of course, I have been leaving the PC alone while encoding, so it strenghens - if not proves - the theory that multitasking and stretching the PC while CCE is running is dangerous.