SMPTE 3:2 film xfred to 29.97 isn’t about extra “frames”, its about extra “fields” distributed so that 4 goes into 5 with no remainder. If film were shot at 30, we wouldn’t have this problem and every frame would be an “A” frame. But we’ve got this ABCD cadence. Where A is the first frame occupying two frames and corresponds to a real 29.97 “frame”. The next frame is a “B” which occupies 3 fields, in other words 1-and-a-half video frames. The real nasty one, the “C” frame, consists of two video fields, “straddling” the third and fourth video frames, and the “D” brings us back to “Doh” by filling the last three fields of the 10 field sequence by straddling the fourth and fifth video frames. So the only way you could get a nice de-interlaced, jitter free still would be on either video frames 1,2, or 5, since they are uniformly and interlaced A, B, or D frames. Editing on a “C” frame is the killer, because that video frame will either contain a field of the B or D frame. Now do you get where the problems might start adding up?
About 10 years ago, I got a look at one of the first plasmas, brought into a post house on spec/demo.
The sales guys asked us to put up some footage to demonstrate how fantastic the new pictures were, and… oops. First tape was a (protection clone) master of a nationally broadcast show, which jumped and showed jumbled fields at every edit. The reason? 3:2 pulldown. The show was shot super16, xfrd at 23.98 A-Frame to 29.97 and cut in 29.97, just like everything else was up until recently. So of course the cadence was totally hooped. An editor has about a 1 in five chance of getting it right, technically, but they edit for story NOT frame numbers. So it was a bit of a surprise for all concerned that what had passed an international QA was generating all this garbage. Of course the plasma could not possibly be at fault — it was brand new of course.
Generally speaking, anything shot 23.98 should be edited at 23.98 and then streamed out all in one piece so that the cadence is correct and uninterrupted throughout the media. This is only logical, from an engineering standpoint. There are other examples of flawed thinking, for example, sending out 23.98 footage on 29.97 tape “dailies” for editorial. That one just boggles my mind. Why don’t you just shoot yourself in BOTH feet?
Now that we have so many flavours of 3:2 cadence, the issues have become much more complicated, of course. Not that I am particularly fond of AP, but it is rather clever and lends itself to a very neat redundancy removal algorithm. But apps like Shake can’t deal with it because it is only able to recognize SMPTE 5:4 spread 3:2:3:2, or 2:3:2:3 if it was A-Frame. So, like all kluge fixes, you should either deal with it at the source, or prepare for pain.
jPo
This IS my blog!