Put simply, my recommended technique is thus:
- demux the audio, and slow it down to 960/1001 (~95.904%).
- write an AVISynth script that resizes the 576 vertical resolution down to 480, and changes the framerate to 23.976 fps.
- encode the script output with CCE as an NTSC DVD - this will make an MPEG-2 file with 2:3 pulldown flags inserted to give an effective framerate of 29.97.
- mux, author, etc. as usual.
There is an alternative method which involves keeping the framerate at 25fps, and then using DGPulldown to set flags to output 29.97fps. This means no adjustment to the audio is necessary. I would check out Doom9 for details.