Re: Color Reference of NTSC Formula?

[Mark McDougall <msmcdoug@no.spam.iinet>]

Bryan Parkoff wrote:

> It is only a question why Deep Red appears in the left of this pixel. The
> answer might be two zero bits between one bit because of no luminance.

Here's what *I* understand about colour artifacting, although I've never 
studied the specifics on the Apple 2 in any detail. So take this with a 
grain of salt..

As you know, colour (chrominance) is encoded as a phase shift from the 
colour burst reference modulated on top of the monochrome signal (luminance).

In an ideal world, this colour component might consist of a sine wave whose 
phase can change instantaneously for each 'pixel', and by any multiple of an 
infinitesimal amount. Naturally, circuits have bandwidth limits as does the 
transmission spectrum so the phase changes have a finite 'resolution' and 
the phase can't actually change instantaneously. Obviously, these 
limitations still allow a reasonable quality picture to be displayed.

The Apple II (as did other computers of that era) generate what should be 
analogue video signals using digital approximations. Indeed, the very reason 
computers have discrete 'pixels' is a by-product of this fact, whereas a TV 
picture raster line has no such horizontal delineation.

You can, for example, crudely approximate a sine wave using a simple square 
wave of the same frequency. If that square wave is passed through a low-pass 
filter, the higher frequency components are filtered out and the resulting 
output more closely resembles a sine wave.

Now, you can't change the phase of a sine wave using a square wave of the 
same frequency. But if you chose, for example, a frequency 4 times higher, 
and approximated the sine wave using 4 consecutive 1's followed by 4 0's, 
then the resulting square would be exactly the same, but you can now vary 
the phase by +/- 45 degrees by inserting or removing an extra 1 or 0 into 
the stream.

It gets more complicated when you start moving away from 4 consecutive 1's 
and 0's. For example, if you toggled 1's and 0's every two clocks (rather 
than 4), then you'd think that you've simply doubled the frequency of the 
colour signal. However, colour is encoded as a phase shift - it's not 
frequency modulated - so that, and the fact that the decoder is band limited 
- means the decoder 'sees' the 'double frequency' as a constantly changing 
phase. You'd no doubt end up with groups of repeating pixel colours.

Also, the resolution of your 'clock' also limits how *quickly* you can 
encode phase changes. If your stream changes from 0 to 1, the value is held 
at 1 for the entire pixel, and the decoder can't 'see' how the waveform is 
going to vary in future, so your next pixel is limited in some way by the 
colour of the previous pixel. Depending on your clock resolution, it may 
take 2 or more pixels to get from 1 colour to the next.

I don't know the specifics of the frequencies involved on the Apple 2, but I 
*suspect* the artifacting is a result of being able to change the phase of 
the signal by +/- 90 degrees only? Can anyone confirm?

Hopefully I haven't sold you a crock of sh*t here... I'm pretty sure it's 
the gist of the mechanism if not 100% accurate.

So if you want to understand how to 'emulate' artifacting I think you need 
to understand both (1) how colour is encoded on NTSC/PAL and (2) how the 
apple generates the video signal.

If anyone knows better, please chime in!

Regards,

--
|              Mark McDougall                | "Electrical Engineers do it
|  <http://members.iinet.net.au/~msmcdoug>   |   with less resistance!"