[Coco] Artifacting (again)

[Ciaran Anscomb]

Mark McDougall wrote:
> On 3/09/2010 5:11 PM, Ciaran Anscomb wrote:
> 
> > The way I do it is with a 5-bit lookup table, indexed by the value formed
> > by the bits two pixels before to two bits after the pixel being considered.
> >
> > So for say, the 8th pixel across, the number formed by the 6th to 10th
> > bits of the scanline indexes into the colour table.
> 
> Interesting. Looking at your source, it's a straight look-up and therefore 
> not surprisingly I am unable to deduce any of your logic for it.
> 
> Would you care to enlighten me (us) on your algorithm? Does it have any 
> theoretical basis, or is it based purely on observation?

For what the colours produced are?  Pure observation.  Of the output
of MESS!

The general theory though is that as the dot clock is an exact multiple
of the colour subcarrier, the frequencies of high-to-low luminance are
easily seen as colour information in a composite signal where both parts
are extracted from one signal.

Looking at a chart of the NTSC I/Q axes, I suppose it's the I axis
that high-low transitions affects most, as that pushes out to red-ish,
or cyan-ish.  AIUI, for each axis, the phase determines its "value",
so as high-low-high is 180deg out of phase with low-high-low, you get
the two ends of that axis.

http://en.wikipedia.org/wiki/YIQ

Disclaimer: I am not a broadcast engineer, I just hang around with them
in the pub.

..ciaran