[old_tv_nut]

Slight correction about PAL chroma phasing:

The B-Y alternates from line to line due to the subcarrier frequency (plus has a slight offset due to the 25Hz frequency offset, which makes it slip phase slowly from the top to the bottom of the screen); the R-Y, due to "Phase Alternation by Line," does NOT alternate 180 degrees from line to line, and has only the slight diagonal offset due to the 25 Hz. This consistent phase should be very helpful in recovering the chroma from the subcarrier dots, I think.

The R-Y line-up can be analyzed in frequency domain also (perhaps making your brain hurt!) as modulation by a half-line-scan-rate (15625/2 Hz) square wave, which suppresses the original color subcarrier and sideband frequencies and creates double the number of sidebands, which are offset by +/- 7812.5 Hz. This can be shown (in either the frequency or time domain) to show that a simple one-line-delay comb filter will work for B-Y, but not for R-Y, since the R-Y dots line up just as though they were luminance stripes (except for that 25 Hz frame offset, of course).

You can draw some really pretty pictures of the 3-dimensional spectra (horizontal, vertical, and temporal frequency axes) of NTSC color and PAL color which show just what is necessary for comb filtering. but I'm not going to try that here. These also show all the flickering cross-color aliases for the two systems and show why some of the PAL crosscolor is more noticeable.

Edit: I forgot that the subcarrier frequency for PAL is also offset by a quarter cycle per line period to prevent the vertical line-up of the R-Y. Makes the cross-color not as bad as it could have been, but still a problem for comb filtering.