View Full Version : CCE hang reason found
24hourloop
26th January 2002, 16:38
Ok,
I found why CCE occasionally hangs with DVD2AVI and the reason is an incorrect ECL that DVD2AVI generates.
I used one movei that consistenly hung, predictably and repeatably. I recreated the hang on Win2k and WinXP.
I then proceeded to run the encoding outside of DVD2AVI using DVD2AVI's ECL - Repeatedly and consistently the encoding hung.
I then proceeded to modify the AAZVS file by adding ResampleSize(44100).
I then loaded the AVS file into CCE and set my parameters manually, however within the same constraints, i.e. mix, avg. and max bitrate. I left the other defaults with the exception of 'Progressive frames' and 'Zigzag scanning order'.
CCE encoded without a hitch. I then proceeded to continue to use DVD2SVCD at step 31 and generated the proper bins.
I believe this is proof that DVD2AVI generates a bad ECL file. I have not been able to discern what the differences are that are causing it as they are too numerous to determine.
My suggestion would be to have DVD2AVI generate a more 'straightforward' ECL file that doesn't play certain tricks on CCE which will solve the whole issue.
I can provide ECLs if necessary.
dvd2svcd
26th January 2002, 16:49
dvd2svcd does NOT create an invalid ECL file! Maybe your test works, but it is NOT the only solution. I have tried what you suggested with ResampleAudio, and sometimes it works sometimes it doesn't. And if you want to know why I know that the ECL isn't corrupt is because I let CCE generate the ECL file. Look closely at what I do when executing CCE. You'll see that I load the avs files and then save the ECL file from within CCE. Then I load the ECL file, change the bitrates and save it again, then reloading the ECL file in CCE. I would like you to compare dvd2svcd's ECL with the one you generate. If you find anything (which I doubt you will as I have done so repeatedly) let me know.
24hourloop
26th January 2002, 18:26
Let me back off for one second: I noticed that DVD2SVCD was set to four passes instead of three. I am running the comparison right now.
A difference I noticed was that non_linear in my file was set to 1 and in yours to 0. (Actually 1 seems like a better choice according to the CCE manual).
I had upped both bitrates (DVD2SVCD and mine to 2520).
Your quality_prec was set to 17 and mine to 40.
Your filter_val is 2 and mine is 6.
I am redoing with three passes now and let you know.
If these settings are indeed causing CCE to fail there definitely is a bug in the encoder and one would have to work around it.
Anyhow, none of this is meant as an attack on you (I very much appreciate what you are doing). I merely try to get to the bottom of this and I believe I am extremely close.
I would say that I know more in six hours.
dvd2svcd
26th January 2002, 19:02
Yeah, you might be close to the solution of that particular movie, but I gurantee you that it is not an allround solution. And as far as I can see you're repeating all the tests I have done too with my faulty movie. Anyway, just don't lay it out as an one and only solution, because it is not, it is only a possible solution to some movies.
24hourloop
26th January 2002, 19:54
Maybe so. This is what I can state so far:
I have done about 80 movies with CCE and never had a hang. I used DVD2SVCD aboutsix times and had 2 hangs. There must be some relationship, don't you think?
What is your theory? How can a concerted effort of all users testing certain aspects shed more light on this?
dvd2svcd
26th January 2002, 20:55
I believe that the problem is YUY2 and too close Avg. and max. bitrate in CCE. YUY2 definately plays a major role in this somehow. Because if I use VFAPI (converting the avs to an semi avi file), CCE never crashes. Also if I use ConvertToRGB24 as the last line in the AVS it also seems to work every time (however the latter I haven't tested intensively). I bet that the 80 movies you did in the past, was either done using VFAPI or that you had a bigger difference between the max and avg bitrates.
Other parameters also seem to have an effect. For instance, not using Auto on Intra DC precision in CCE can have a positive effect, also using 0-255 in the Luminance Level also seems to work on some movies. Unfortunately these two things is not stable and it doesn't work on all movies. And also your suggetion about using ResampleAudio seems to somehow solve other problems too (especially on Athlon computers)
24hourloop
26th January 2002, 21:52
I would say the last 20 movies I did used MPEG2DEC, I always stayed in the YUV domain all the way. I always used the TV luminance scale.
I never changed anything in the CCE settings besides
a.) VBR - 3 passes
b.) No audio
c.) Avg (generally between 1900-2100), min 0, max 2520.
d.) Progressive audio, zig-zag scanning.
When CCE hangs task manager shows the process looping. I am not that deep into debugging and windows but it sure would be interesting to see whether it is any DLL.
It is too bad that this can't be checked under 2.62 whether the symptom persists due to the AVS check they apparently put in.
chainsaw135
26th January 2002, 22:27
I have been hoping 2 find a solution for the CCE problems.
But i'm not sure there is really 1 easy fix. I believe there is mulity problems stemming from the CCE.. 2 peoples hardware/software . And 2 fix all of those would be near imposible.
I have had CCE crashes but no way near the same reasons of why others have, not sure how many people even realize why they are having them. I found in my case it was 100% of the time would crash when I adjusted my VIrtual Memory 2 low if i allow windows 2 set the size everything would work fine ....I'm not saying this is the fix heck i dont believe there will be a 1 fix suits all type thing here...but you can always hope for one:)
24hourloop
26th January 2002, 22:46
Never had a single crash, ever. (If you don't count the CRC errors).
24hourloop
27th January 2002, 00:02
Ok,
I redid the movie with VBR 3. It hangs again. The only differences between the ECL files are the ones I outlined.
vBulletin® v3.8.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.