Conditional filtering w/ TFM (futurama PAL)

[check]

I recently went out an brought the Australian PAL release of futurama (region 4). The quality is very nice (no blending, yay!), but I'm still running into a few problems in the deinterlacing.
The stream seems to drop random fields, which seems to throw a spanner in the works.
Here's a sample of what I mean. The scene where leela lunges towards the camera seems to switch backwards and forwards, and TFM can't seem to get it by itself. http://cccp.project357.com/index.php?dir=samples/&file=futurama_r4PAL_-_fieldorder_woes.vob

I've found that I can mostly fix the problem by running TFM through conditionalfilter where if the output is still combed, the field parameter is switched:
DGDecode_mpeg2source("G:\Video Editing\FUTURAMA_DISC_1\VIDEO_TS\ep1.d2v",info=3)
a=last
try1=TFM(d2v="ep1.d2v",slow=2,mode=4,pp=1,field=0) #.Tdeint() #,order=0,field=0
try2=TFM(d2v="ep1.d2v",slow=2,mode=4,pp=1,field=0) #.Tdeint() #,order=0,field=0
ConditionalFilter(try1,try2,try1,"IsCombedTIVTC()","=","true")
Tdeint(hints=true)

This seems to work pretty well, but there are plenty of places I'll have to use a manual overrise file for. Can anyone offer suggestions on a way to go about this that produces a more accurate result?

[scharfis_brain]

this is a DEFT conversion from NTSC to PAL.
the scene where she runs towards the camera at the end of this clip shows a lot of blended fields. That is typical for DEFT conversions.

[check]

d'oh, well after realising I had messed up the script:
try1=a.TFM(d2v="ep1.d2v",slow=2,mode=4,pp=1,field=0)
try2=a.TFM(d2v="ep1.d2v",slow=2,mode=4,pp=1,field=1)

After that it works much better >_>

Can you point me to a definition of what a DEFT conversion is? Is it just a 'normal' NTSC->PAL conversion?
I played around with restore24 on that scene which seems to make it worse. I remember a thread from a while back titled something along the lines of "state of the art in 60i -> 24p" - would solutions in there be a good avenue of investigation? If not, what sort of filtering should I look at to help repair this?

[scharfis_brain]

DEFT is the abbr. of Digital Electronic FIlm Transfer

this means the conversion machine searches for a 3:2 pattern in the source. if it finds this pattern, it will convert it to a 2:2 pattern. everything else becomes blended.
this machine puts out 48i. It will speed up afterwards to 50i

EDIT: corrected typo

[Mug Funky]

don't you mean 50i ;)

i don't mind these conversions - it's a good compromise really.

...of course we can't afford that type of equipment where i'm at, so i have to use avisynth to do it :). it's not an easy problem, let me tell you. but the good news is to get a progressive backup you usually only need a smart deinterlacer to get a fairly good result.

[scharfis_brain]

DEFT conversions aren't ideal at all.
they blur like hell in vertical, even if they found telecine and convert the 3:2 pattern to 2:2.

early deft converters introduced stairstepping due to vertical field misalginment during upsizing from NTSC to PAL.

the list of issues can be continued further.

At least these converters offer motion fluidity for FILM footage.

[Mug Funky]

ah, i meant my home-brewed emulation of it. also not ideal, but _almost_ at the point where its output rivals anything i've seen on broadcast TV (ie motion-compensated conversion, IVTC, speedup, blend or changefps, etc). only things missing are intra-frame operation (i'm not going to bother), and good 30p detection.

i've never dealt with one of these machines before, except for their output (several warner DVDs appear to have been done this way. not too much vertical blur, but definitely some field misalignment, and sometimes some remaining interlaced chunks within a frame - take a look at A Clockwork Orange).