Here's an example and explanation on how to set up a 8-bit color palette on the VS23S010 in an efficient way.
VS23S doesn't have a color look-up table, or "palette memory" as it's often called, because adding one to VS23S would have been prohibitively expensive. Instead, it has a versatile micro-code engine that picks bits from the video memory and assigns those to Y, U and V coefficients in the video modulator. Here's one example of usign the microcode engine in an interesting way to make a nice 8-bit palette (256 colours).
Assigning bits from pixel color value to color coefficients
8-bit palettes are attractive from programmer's point of view, because 8-bit colors will have separate pixels in separate bytes (1 byte per pixel), so each pixel is uniquely addressable in the framebuffer memory. When VS23S010 is drawing the picture, it will load pixel data from memory into a shift register inside the microcode engine. The microcode engine picks bits from the shift register into color coefficients Y, A and B. Y bits modify luminance, e.g. set the pixel's brightness. A and B commonly modify the V and U components in PAL YUV colorspace, or Q and I components in NTSC YIQ colorspace (depending on the phase value of the color burst).
A clever and surprising feature of the microcode engine is that bit assignment and shifting are done independently from each other, e.g. bits of the pixel color value can be assigned into more than one coefficient. For example, see the assignment below:
- path3220-1.png (8.08 KiB) Viewed 14459 times
Below is a picture of the palette, photographed from a modern LCD television:
- VS23S010 TV-Out Palette Test.jpg (84.83 KiB) Viewed 14438 times
With the burst value of 0xB5, the microcode that forms this palette can be seen in the picture: "B62, A63, Y33, N10", which reads out:
"- For B coefficient (u) pick 6 bits, then shift out 2 bits.
- For A coefficient (v) pick 6 bits, then shift out 3 bits.
- For Y coefficient (luminance) pick 3 bits, then shift out 3 bits.
- For the final step, pick 1 bit for No coeffient and don't shift any bits."
From the picture below, you can see how this microcode correlates to the bit assignments:
- path3220-1.png (8.08 KiB) Viewed 14459 times
Colours on the VS23S are not limited to 8 bits, or even 16 bits; the maximum color depth is 20 bits per pixel (8 bits for Y, 6 bits for U, 6 bits for V). But 8-bit or 16-bit colors are nice because they allow individually addressable pixels.
PS. One final note. In the picture above, also value of d139 was written to register 2d, making the line count odd. This, together with the line length setting, has the effect of making the color carrier be in opposite phase in successive frames, which helps the TV to separate color and luminance. That makes the picture better.
-Panu
Really, this is the biggest issue with using your chip for applications that need RGB equiv. A few percent off is really not noticeable, and you can really tweak the microcode to get under 1%. I really wish this program would have been available a year ago!