Improving the Colour Test code for analog VGA

[cocoacrumbs]

Hi,

On https://projectf.io/posts/fpga-graphics/, there is this verilog code (Colour Test section):

logic [3:0] paint_r, paint_g, paint_b;
always_comb begin
    if (sx < 256 && sy < 256) begin  // colour square in top-left 256x256 pixels
        paint_r = sx[7:4];  // 16 horizontal pixels of each red level
        paint_g = sy[7:4];  // 16 vertical pixels of each green level
        paint_b = 4'h4;     // constant blue level
    end else begin  // otherwise black
        paint_r = 4'h0;
        paint_g = 4'h0;
        paint_b = 4'h0;
    end
end

On analog VGA this might cause troubles if you want to display anything else than black outside the colour square. When I tried green instead of black I got very weird results (the colour square suddenly only showed shades of red, with horizontal bars of shades of green next to it).

I finally figured out that the analog R, G, B signals have to be black when outside the display area (de == 0) and I came up with this code which will properly show a cyan color (instead of black) outside the color square.

    always_comb begin
        // colour square in top-left 256x256 pixels
        if (sx < 256 && sy < 256) begin
            vga_r = sx[7:5];                      // 8 horizontal squares of each red level
            vga_g = sy[7:5];                      // 8 vertical squares of each green level
            vga_b = 3'h0;                         // constant blue level
        end
        // paint other pixels in cyan
        else if (de) begin
            vga_r = 3'h0;
            vga_g = 3'h7;
            vga_b = 3'h7;
        end
        // don't output anything when display isn't enabled (important!)
        else begin
            vga_r = 3'h0;
            vga_g = 3'h0;
            vga_b = 3'h0;
        end
    end

It could be that the behavior I see is dependent on the [LCD] monitor I use (a very old Sony LCD monitor with 4:3 aspect ratio (16:10 didn't exist back then)). Maybe that more modern ones don't have the problem I observed. Interestingly, using the DVI input on that same monitor (using a ULX3S with the original code and HDMI to DVI cable) doesn't show this problem.

[suarezvictor]

I had same issue, it seems black is "lowest análogo level" and not zero (the signal is DC filtered), see some images here litex-hub/litevideo#42

[WillGreen]

Hello, and thanks for your message. You’re right that many VGA displays are confused by colour data during the blanking interval. I briefly cover this in the blog post section on Arty VGA output. All the Arty examples apart from the colour test make this explicit, for example, the Swedish flag.

I’ll amend the colour test design to clarify the distinction between background and blanking colours.

[cocoacrumbs]

Thanks for the feedback. It took me awhile to realize what was going on and first thought that something was wrong with the simple_480p code :-( Hence, I decided to report my experience so that other newcomers could know what is happening if they would stumble into the same issue.

[WillGreen]

I was planning a quick fix, but on reflection, I realised I should do this properly. This new approach ensures the colour channels are always zero in the blanking interval, freeing the user to mess with the painting colours as they choose.

You can see the revised designs in:

• FPGA Graphics
• Racing the Beam

And take a look at the updated blog post: https://projectf.io/posts/fpga-graphics/#12-bit-colour

Let me know what you think, @cocoacrumbs @suarezvictor, I plan to apply the better blanking to all designs over the coming weeks.

[cocoacrumbs]

I agree, your solution is a better approach to this problem than my quick hack in my opinion.

[WillGreen]

I have applied better blanking to the remaining FPGA Graphics designs.
I'll update the related blog posts soon.

Designs in projf-explore/demos will be updated soon.

[cocoacrumbs]

Thanks.