Date: 17 Apr 97 09:53:23 -0700
Subject: Re: Apple IIc//IIe graphics modes.
From: "Eric Jacobs" <no@no.no>

Aw.. don't shoot down the graphics design of the Apple II! That's the 
best part of the whole computer! Its insanity makes it my personal
favorite, I'll tell you that. A video system built solely using discrete
counters and adders.. that's pretty impressive.

A lot of the reason the hires color system is so confusing is that it's
not built on an RGB color model, but directly on the NTSC composite
television standard. Most other computers (including the GS) generate
the image in RGB, and then display it. (When systems like the GS and the
Super Nintendo need to output to composite, they encode the RGB signal
using a matrix chip, and then send it out.)

The original Apple II, on the other hand, directly generates a composite
signal. In an NTSC video signal, the color (chrominance) signal "rides"
on a 3.579545 MHz subcarrier. The subcarrier is then decoded and
referenced to an internal oscillator in the TV or monitor, which
determines its amplitude and phase. The amplitude is the color's
"saturation": how much color there is. In the Apple, this is always 0,
25%, 50%, 75%, or 100%.

The other element is the phase. The phase specifies an angle in an
imaginary color circle. Thus if the phase of the chrominance signal is
exactly in phase with the color reference, it is 0 degrees-- the color
is yellow-orange. 180 degrees is blue. 90 degrees is red, and so on.
The internal oscillator is kept in sync by a signal called color burst
in the horizontal blanking before every line. Using the tint knob on
the TV and the variable capacitor on the motherboard fine tunes the
color reference.

Any 3.579545 MHz component in the video signal will produce a color on
the screen. The Apple II text and hires modes clock the video signal
at 7.15909 MHz-- twice that of color burst. Therefore, the bit pattern
10101010... will produce a color on the screen. The only other color
produced that way is its complement, 01010101... In the original Apple
II design, these two colors were purple and green (respectively). 
Combined with black and white (which have no color component), this 
gave the original four colors.

After Revision 0 of the Apple II motherboard, additional circuitry was
added to produce two more colors. When bit 7 of the data byte was 1,
a flipflop was selected that would delay the bit stream by 70 ns (that's
90 degrees). Thus, the Apple could now generate six colors, black, 
white, purple, green, blue, and orange (135, 315, 225, and 45 degrees to
color burst, respectively). The only limitation was that this new
palette select bit applied to the entire byte. So it was not possible
to display blue and green in the same byte, for instance.

Now, for Nate's question about colors spanning bytes with different high
bits: if you're going from D7=0 to D7=1, the flipflop holds the bit for
an extra 70 ns. This creates a bit pattern that's a different color
entirely: in fact, it is a 25% or 75%-intensity color from the lores
palette! For example, going from green to orange will give you yellow;
purple and blue will give you light blue.

This same phenomenon is also responsible for the Mysterious Pink Line.
This is a orange-pinkish line that appears under certain conditions
along the left edge of the screen. It occurs when bit 7 of the left edge
of the screen is set, and bit 6 of the previous byte is also set. The
flipflop holds that value of bit 6 during the first 70 ns of the first
byte of the new line. The result is an eerie pink half-dot at the left
end of that line. For example, for scan line #1, set $2080 to $80, and
$207F to $40. $207F--an unused screen location (screen hole)--is
actually being used to store that dot!

Sadly, the Myserious Pink Line is not correctly emulated under the
Apple IIGS's video chip. It seems that Apple's engineers had the sudden
thought that maybe the Apple II video modes should make sense.. i
dunno.

Actually the Apple II's video could have been much better if Apple had
adhered closer to the NTSC video standard. For example, the number of
cycles of color ref in one horizontal line should be 227.5, not 228.
The number of lines per field should be 262.5, not 262. (Apparently,
Woz much preferred integers.) Although these deviations seem small, they
actually would have fixed many of the color anomalies, including 
virtually eliminating that annoying color fringing. Just take a look at 
the Nintendo or Super Nintendo's color display.. it's absolutely 
beautiful. There are no wacky color fringes on those machines. Or look 
at the credits to a movie on your VCR. That text is no bigger than an 
80-column display, and it looks great.

Ah, the wonders of composite video.. now how can the GS and its dumb old
SHR compare to this??   :)
-ej