One more Cineframe24/HDlink 24 question

Laurence wrote on 11/4/2005, 3:35 PM
I was thinking about the cadence of Cineframe 24 footage and how it is not exactly 24 evenly spaced frames per second. Then I realized that 24p with a 3:2 pulldown isn't exactly 24 evenly spaced frames per second either, and nobody has ever complained about that. Then I was thinking that 24P which originated from CF24 might actually be a closer match to the timing of interlaced playback of 24p footage and maybe the timing discrepancies I think I'm seeing aren't really that big a deal after all. The timing on the regular TVs that most of us still use might even be better than true 24p. Is this the case or am missing something?

Comments

Serena wrote on 11/4/2005, 5:18 PM
Not entirely sure I've understood you correctly. The process for generating 60 fields from 24 frames works well, so it can be expected to look OK. Going the other way requires some fields to be discarded and other fields from adjacent video frames to be averaged (smoothed) to minimise motion artifacts. Done well it works OK and in effect there are indeed equally spaced time slices. Various people consider the Sony Cineframe 24 method to be flawed, but not having viewed this I cannot comment.
The question, I think, is how does it look to you? Or if you don't like it now, can you get used to it? Like people accept sharpening artifacts on video?
I haven't gone back through the threads to check your interest in Cineframe 24, but if you're not releasing on film then is there need to use 24fps? If you're wanting progressive to remove interlace artifacts, other approaches can be explored (as you have been doing).
As set up in the Sony menue, cineframe 24 gives you both cinegamma and 24fps. You can change the profile to use cinegamma with 60i, should that be what you would like.

I find interlace effects in HDV to be unacceptably distracting and there is no way I can accept what I see on my 3 metre screen (a quite tolerable problem when downcoverted to SD). All the arguments in favour of higher temporal resolution etc for 60i compared to 24fps are insignificant in comparison. Up to now the only way I've been able to get acceptable images is by shooting Cineframe 25 (no motion artifacts and the offer of an extra stop in low light although with a loss of vertical resolution) or rendering from Vegas at 24fps (or capturing in Cineform at 24fps). Clearly there are other things to be tried (as were mentioned in your previous thread).
farss wrote on 11/4/2005, 6:32 PM
No you haven't missed much. Trying to present 24fps is difficult, that's why projectors have two blade shutters, at lower frame rates they go to 3 blades. This means the eye is flashed with the same image several times to reduce the flicker.
That however doesn't mean that 24p with 2:3 pulldown isn't true 24 fps though!
To get a true progessive image out of the Z1 is I guess technically impossible without loosing resolution and even then you've still not getting a true filmic experience as the shutter speed is too high. I don't know just how much that'll upset the viewing experience but even on regular video I find fast shutter speeds on fast motion quite distracting. Still in this case the shutter speeds are still close, not like some broadcast live ocverage that seems to be shot at 100ths of a second just to get good SloMo!
Perhaps the answer to the riddle lies in shooting either 50i or 60i and getting very good de-interlacing to 25p or 30p. And that I think is the crux of the issue, most de-interlacers are taking a pretty simple approach. I had somewhere here a manual for the Leitch standards converter. Part of the conversion process involves de-interlacing and as this is broadcast kit the results have to be spot on and certainly they're not going to be throwing resolution away.
Now what this box uses is adaptive motion compensation. In other words it moves objects within a field so they line up with the previous field. Now I'm pretty vague on how this works, I've seen a BBC paper from some time ago on how this is done but the math lost me. I am told that some of the high end software tools that are perhaps afforable also can pull off this trick, tools such as Twixtor and Digital Fusion however when I did look these were only for SD and / or the HD versions cost very serious money.
So I think the tools are out there to do this job very well, at the moment they would seem to be very expensive but if there's enough uptake from HDV users then the price has got to come down.
That might change though with the next generation of HDV cameras hitting the market soon that will shoot true progressive scan.
Just one note about adpative motion compensation, the one thing acording to the manual that spins it right out is noisy video, perhaps that's why no one's offering it to this part of the market.
Bob.
MH_Stevens wrote on 11/4/2005, 8:12 PM
Not wanting to be too tangential, but why do video cameras shoot interlaced anyway? Is it technically easier with slow CCD's or is it because most footage ends up on a TV?
farss wrote on 11/5/2005, 2:57 AM
From my fairly limited knowledge of how these things work CCDs are scanned and that takes time. Scanning half the elements and then the alternate elements makes for CCDs that are easier to build. JVC in their response to the split frame offset problem on the hd100 explained that the CCD they use is scanned as two halves because trying to scan such a small CCD quickly enough would create so much heat as to make it cook.
Canon have taken another approach with 24F, the CCDs are 'exposed' once, scanned as two fields and the two fields merged to yield a frame.
One question I've yet to have the answer to is why are cameras with smaller CCDs cheaper, everything I've read from lens design backwards through the camera seems to say that bigger CCDs remove many design issues so a camera with a single 1" CCD should be cheaper than one with 3 1/3" CCDs.
Bob.