Got quite cryptic ? But I guess many people out there understand well, what I'm talking about, as too many have similar problems.

To start with: I have PVR-250 and have captured only from S-vid input (if it makes difference) and in DVD and SVCD modes only and only from sat. Problem is, I havent managed to get single PERFECT disc so far. BTW all recordings burned so far were captured under 98SE.

Lets take SVCD first. Single problem - gradual unsync on burned CDs. Captured video is allways in perfect sync. SVCD hardly scrollable in Ulead DVDMF (hangs often after going past 15 min mark), but DVD not so bad. Optional trim in nanocosmos - perfect sync. Edit and re-encode in Ulead (from DVD format) - still perfect sync. Burned disc - out of sync. Remux, cut, join in TMPGEnc or BBmpeg at any stage produces out of sync video. IIRC only re-encoding with Ulead for SVCD and burning with Nero instead gave in-sync disc. But I can't re-encode everything, right!

Now to DVD. Here things get much more complicated. No gradual unsync problem as such though. Havent made simple capture-burn discs either (I guess there would be no problems). But cutting out undesired content seems to be nearly impossible task. Just example which illustrates all of my problems:
Having capture in 3 files (4GB cap!). Marked down all cut points (I) from DVD Workshop. Got 8 parts to cut-join. Gave nanocosmos a try. Resulting video had awkward transitions and refused to scroll after some point (15-17 min in 1+ h vid). It actually scrolled anywhere, but jumped back to this point just after releasing slider or hiting pause(when playing). Now TMPG ... in part 5 it found droped frames and went no further. (later between parts and in part 8 too). After that I cut vid down into pieces and removed corrupted GOP-s with nanocosmos. All parts are playing well and range from 3.5 to 16 min in size. At this point I left original vid alone.
Now I was trying to merge clips. Nanocosmos gave same results as before. Different "scroll-stop" point though. TMPG went all the way this time. There was 0.05 sec diff between vid and audio length. Result: in part 5 audio gets out of sync badly (noticeably in beginning and 1 sec or more through the end of 7 min clip), part 7 plays half-speed (jerky) and part 8 refuses to play at all.
Merging 3+4+5 - perfect sync. Merging 1+2+3+4 - perfect sync! Merging 5+6+7+8 - also perfect!!! Merging 2 latest to get full length - exactly same as before.
Gave DVD MF a try (was desperate enough to re-encode). Somewhere in part 7 it stoped decoding and encoded same frame over and over till the end. Same in part 8.
Finally burned in 2 parts - plays well, but should be in 1 piece! And transitions are blocky.
What have I done wrong?

Problem is, this vid is not the last one of its kind. I have allready marked down droped frames from my next vid to edit and am fearing worst. If problem IS droped/corrupted frames at all. TMPG found 1 corrupted I and 1 corrupted P for me (respectively 16 and 4 frames have same picture in DVD WS). Didn't run ReMPEG on this to check GOPs, and am away from my home PC for couple of days I guess, there will be couple of corrupted B-s too.

My questions are
1. How to make working SVCD from PVR capture?
2. How to cut off undesired content without generating sync issues?
3. How is it at all possible, that video and audio, having same length in samples*44100 or 48000/frames*25 are geting out of sync?
4. If it is droped frames, how can I fix this? (players somehow can)
5. Once again, how can I fix droped frames I can notice, without cuting some frames out and/or re-encoding full strip and geting "jump" in audio?
6. Is Ulead's "Smart render" any good? It is as slow as re-encoding and that's, why i fear, it is not too smart at all (VS6 Trial). As I understand, it should render only transitions, as all GOP sizes are allowed (right?) and leave other parts unchanged. And this should be much-much faster.
7. How to merge clips and get good transitions (like with VCR) at least on I-frames.
8. Will migration to Win2k/NTFS make any good (except cap size).

If someone, who was able read this to the end , has good suggestions, i'm happy to give them a try.