Quote:
Originally Posted by Electronic M
From what I understand kinnescope color recovery has only successfully been done on PAL color kinnescopes. PAL has the advantage of the chroma subcarrier phase reversing every line so phase (hue info) can be recovered by comparing adjacent lines relative phase.
NTSC keeps color carrier phase constant so without the sync burst there is no relative reference like what PAL gives.
Can NTSC kinnescopes practically be recovered using this method?
|
You are correct. Areas that have significant energy on the axis that is in phase from line to line can be a reference. Then it is a matter of resolving which quadrant the chroma is in, which is relatively easy.
The people who did the PAL recovery have surmised that it would be impossible for NTSC.
I still strongly believe it could be done, but in the NTSC kinescope, you have several compounding (and confounding) issues.
1) The dot phase changes 180 degrees as you go from one line to the next in a FIELD. Since the film frame contains two fields, the dots in two vertically adjacent lines in the FRAME are the same phase. The vertical "switching" of phase means that the geometry of the recovery scan must be corrected to match the geometric distortions of the original image within very tight tolerances. Any vertical jitter must be compensated. I think the recovery scan has to be done with a pull-down mechanism rather than continuous film motion, since continuous scans can have "Jello" effects if the film has variable shrinkage or doesn't move at an absolutely constant rate.
Working on this at home, I was only able to set overall height and width, so there was an area near the center of the image that roughly tracked the right phase, but you were 90 degrees off by the time you deviated more than 20% or so away from the center. This resulted in concentric rainbows of hue, getting more closely spaced as they went toward the edges of the image. Also, they were worse on the left side than the right, due to horizontal non-linearity in the kinescope tube.
One hopes that the geometry would not vary hugely from frame to frame, but I really don't know. If the image size breathed due to tolerance on high voltage regulation of the kinescope tube, that might require tweaking of the recovery scan size on a frame by frame basis. I don't believe the BBC encountered this problem, but we would surely have tried to pick their brains if our project went forward.
2) The PAL kinescopes were made at 25 frames per second, so each film frame contained exactly one video frame. NTSC used an inverse 3:2 pulldown, so some film frames had fields mixed from different video frames. Because there is a four-field sequence of chroma phase vs. line number, this messes up the vertical tracking of phase discussed in #1 above.
3) Because the NTSC phase cannot be resolved down to a quadrant decision, the phase adjustment would have to be manually aided. Due to film weave (horizontal placement variation) from film frame to film frame, the phase would also need to be tracked from frame to frame. A way of doing this might be to correct the first frame in a scene by eye (or selecting a skin tone area for automatic adjustment) and then assisting a hue tracking algorithm by pointing to a selected colorful object in each successive frame; or automatically tracking a colorful object with a motion detection algorithm.
So, there was still a lot of experimenting to do at the point where we gave up due to the unsuitability of the film.