Thanks for all the advice,
I took option 3, the source is only 25i from VHS. I just rendered it out straight from the timeline as NTSC mpeg2 but turned on "force resample" and "reduce interlace flicker" and then built a NTSC DVD in DVDA.

Looks OK, would look a LOT better if they'd given me a DV copy to start with.

I don't really understand why they asked fot it as a NTSC DVD, I thought all DVD players can handle either and do the conversion internally or is it just that most PAL TVs will accept the NTSC feed from the player. Maybe they're not so forgiving in the US.

I had asked SoFo just how good the PAL to NTSC transcode is. They said they'd tried NTSC to SECAM and could see difference between the results from VV and a S&W box. Certainly from the fairly crude tests I've been able to run we could see no problems.

We captured PAL from a DSR11, rendered out as NTSC avi and then recorded back to DSR11 switched to NTSC and then played back through broadcast monitor and it looked OK. Probably should run the test again with some very fast motion, sharp edge footage to really check for artifacts.

I don't know if it's of any help to you and I've not used it myself but you could check
DVFilm Atlantis at:

http://www.dvfilm.com/atlantis/index.htm

cheers

Adding to what RBartlett posted, *might* be an interesting something to try with supersampling, or variations of this sort anyway...

There's filters for V/Dub that split the fields into separate frames. Going from 25i PAL, it'd be Slowww going with supersampling interpolating from 50 fps, but I'd think the results might be outstanding...

"I thought all DVD players can handle either and do the conversion internally or is it just that most PAL TVs will accept the NTSC feed from the player. Maybe they're not so forgiving in the US"

According to the FAQ here: http://www.thedigitalbits.com/officialfaq.html - the big deal is the players, those in the US only playing one format, those in PAL countries playing either. The individual NTSC/PAL DVD spec applies, based basically on the formatting encoded in or with the mpg2.

My take, FWIW, & without any docs to back it up, is that an mpg2 is an mpg2. The difference in PAL & NTSC formats is the height, not the width, and the player makes up the difference between 24p & either 29.97i or 25i. If you have a letterboxed or non-letterboxed anamorphic video, as long as the player got the width correct, the height would be irrelevant, as the PAL bar might be higher then that appearing on an NTSC 4:3 TV. For full screen of course the standards would be less forgiving, and those tend I think to obey the specs re: height etc. Would explain why the features added to a DVD are often NTSC all the way. They have to do a different region anyway, so I would imagine a commercial movie release would be encoded widescreen, the vob/mpg2 flag set for either format as nec., and the full screen features localized and added as more of an afterthought sort of thing - the core vobs/mpg2s would not be created twice. Otherwise might be no reason to include multi-lingual tracks on so many US releases, as it would shave a few pennies off the cost perhaps moving the data around prior to pressing. But again, that's just my take.

In theory going to 29.97i is more correct, but in the US some shoot with PAL cameras to get closer to a film look, moving the content to 24p, as it's often less costly then shooting 24p, and not such a big leap as going from 29.97i.. Depending on the feel of the piece, easiest conversion *might* be to just downsample to 24p with a pan/scan to 480 height rather then dealing with quality issues further resampling the vhs source. If you could control the viewing environ, might even get away with a twist on anamorphic, keeping the set height, squishing into a 480 height frame, setting the mpg2/vob height to std NTSC 540 or better. VERY occassionally, going from VHS have captured reduced size to avoid fields entirely, then set the encoder to full frame - when I've been able to use this, the lesser amount of data was compensated for by the coresponding increase in original clarity not having deinterlace blurring/resampling. IF something like that did work, & if you had the leeway, could set the properties to 24p, render as is without resampling, reimport into vegas, timestretch the audio to match, in theory encoding with the only video sampling done by the encoder.