From:
                                                                    ne 12:26
Subject: Re: TECH: RGB Values for Color/Luminance pairs

[posted and e-mailed]

Patrick Maloney (maloneyXXX@merlin.albany.net) wrote:

> With regard to this blue/green problem, I got the AtariWin800 source
> and initial tests indicate I just have to change the palette (as someone
> else indicated).

That was me, I think.  :-)

> Anyone know where I can get info on the appropriate RGB values to
> ues for each color/luminance pair?

That's a tricky question to answer, because the appearance of each
hue/luminance pair will vary depending on the way the tint, brightness, and
contrast are set on your TV, as well as whether you're using NTSC or PAL.
Apparently even the individual Atari can be a factor, depending on how the
potentiometer is set.  I used to adjust my TV settings to get what I thought
were the best hues, based on what the references said they were, and to make
the luminances distinguishable from each other.  I wrote a simple program to
display all 256 colors on the screen at once, using display list interrupts.
Whenever I wanted to adjust the settings on my TV to get the best display for
my Atari, I would run that program so I could see all of the colors at once
while adjusting the TV settings.

Unfortunately, my Atari disk drives died long ago, and I never replaced them,
so I can no longer run my old Atari programs.  Several months ago I bought a
Snappy! digitizer for my PC, for the sole purpose of capturing Atari screens on
my PC so I could determine which RGB value to use for each hue/luminance pair.
I played with the brightness and contrast settings of the Snappy! program to
make each luminance distinct, but I don't think I adjusted the tint.  And
instead of capturing all 256 colors at once, I just captured all 16 luminances
of the hues at once, but I'm not sure if the colors are accurate.  For example,
I remember hue 15 as being brownish, but the screen captures I made show it as
greenish, which isn't right.  One of these days I'm going to play with
Snappy!'s settings to get a better result on *all* of the hues and luminances,
then make a new set of screen captures to work with.

Anyway, I think the hue you're most interested in is hue 10, variously
described in the Atari literature as "blue-grey" (Mapping the Atari, p. 164),
"turquoise" (Atari XL User's Handbook, p. 215; Atari 400/800 Basic Reference
Manual, p. 50; and GTIA Chip NTSC, sheet 4), or "blue-green" (Atari System
Reference Manual, Chapter 8).  The Atari800Win and Stella emulators display hue
10 as greenish, rather than bluish.  All you really need to do is redefine the
RGB values for color 160 (hue 10, luminance 0) through color 175 (hue 10,
luminance 15).  At least, those are the only colors you need to worry about
"with regard to this blue/green problem."  As someone else indicated in another
post, Star Raiders uses color 160 (hexadecimal A0) for the background when
shields are in use.  Thus, if you change that color, Star Raiders on the
Atari800Win emulator should look more like what you're used to seeing on the
Atari.

If you do a screen capture of all 16 luminances of hue 10 (as I did), you can
pull it up in a paint program and inspect the RGB values, then change the
palette in Atari800Win.  That sounds simple, but it turns out to be
complicated.  The screen capture I made of hue 10 contains 5173 unique colors!
When I cut luminances 1 to 15 out of the image, the part of the image showing
just color 160 (luminance 0) still has 432 unique colors!  So I reduced the
cropped image to a depth of 256 colors, optimized to the nearest colors,
resulting in 20 unique colors.  I reduced the new image to a depth of 16
colors, optimized to the nearest colors.  Finally, I examined the RGB values of
these 16 colors, added them together, and divided by 16 to get these averages:
R=0, G=43, B=130.

When I create a color with those RGB values and put it next to the original
cropped screen capture of color 160 (in all of its glorious 432 colors), it
does look like a pretty good fit.  However, the RGB values I came up with
depended on Snappy's settings when I made the original screen capture, the way
I reduced the color depth, and the way I averaged the resulting colors
together.  (I assumed that the 16 colors were used in equal proportions.)  It's
likely that someone else could use this same technique, yet end up with
different RGB values.  And then there's the issue of NTSC versus PAL colors.

Nevertheless, using the technique described above, I came up with the following
RGB values for the 16 luminances of hue 10:

color 160 (hue 10, luminance 00) : R=000, G=043, B=130
color 161 (hue 10, luminance 01) : R=000, G=055, B=144
color 162 (hue 10, luminance 02) : R=004, G=074, B=156
color 163 (hue 10, luminance 03) : R=025, G=085, B=163
color 164 (hue 10, luminance 04) : R=044, G=100, B=172
color 165 (hue 10, luminance 05) : R=062, G=115, B=185
color 166 (hue 10, luminance 06) : R=082, G=129, B=200
color 167 (hue 10, luminance 07) : R=099, G=147, B=214
color 168 (hue 10, luminance 08) : R=104, G=154, B=213
color 169 (hue 10, luminance 09) : R=121, G=169, B=232
color 170 (hue 10, luminance 10) : R=138, G=185, B=246
color 171 (hue 10, luminance 11) : R=151, G=203, B=253
color 172 (hue 10, luminance 12) : R=167, G=219, B=255
color 173 (hue 10, luminance 13) : R=182, G=232, B=255
color 174 (hue 10, luminance 14) : R=198, G=247, B=255
color 175 (hue 10, luminance 15) : R=216, G=255, B=255

Michael Rideout