Our Scan Map for P5,64x32, ICN2037, RUL5158C

[MicroframeDev1]

The following is our mapping for P5 front maintenance panels with ICN2037 led driver chips and RUL5158C line decoder.
This is tested working.

First the human readable version:

struct ScanMapICN2037_RUL5158C {
    static VirtualCoords apply(VirtualCoords vc, int panel_pixel_base) {
        int ysection = (vc.y%32) / 8;

        switch(ysection) {
            case 0:                 //y 0 to 7
                vc.x = (vc.x*2)+1;
                break;
            case 1:                 //y 8 to 15
                vc.x = (vc.x*2);
                vc.y -= 8;
                break;
            case 2:                 //y 16 to 23
                vc.x = (vc.x*2)+1;
                vc.y -= 8;
                break;
            case 3:                 //y 24 to 31
                vc.x = (vc.x*2);
                vc.y -= 16;
                break;
        }

        return vc;
    }
};

Next the speed optimized version. This uses a lookup table to speed up execution:

// high nibble = y_sub / 8, low bit = x_add
constexpr uint8_t ScanMapICN2037_RUL5158C_LUT[4] = {
    0x01, // 0: x+1, y-0
    0x10, // 1: x+0, y-8
    0x11, // 2: x+1, y-8
    0x20  // 3: x+0, y-16
};

struct ScanMapICN2037_RUL5158C {
    static inline VirtualCoords apply(VirtualCoords vc, int /*panel_pixel_base*/) {
        uint32_t ysection = (vc.y >> 3) & 0x03;
        uint8_t m = ScanMapICN2037_RUL5158C_LUT[ysection];
        vc.x = (vc.x << 1) + (m & 0x01);
        vc.y -= (m >> 4) << 3;

        return vc;
    }
};

For these panels I customized the following settings:
line_decoder = HUB75_I2S_CFG::SM5368;
Color channels are swapped compared to labeled pins: RGB -> GBR.

Hope this helps someone.

[board707]

For these panels I customized the following settings:
line_decoder = HUB75_I2S_CFG::SM5368;

Customizing the line decoder code is a wrong approach. Line decoding type and pixel mapping are independent setting and must be tuned separately.
What is the scan factor of your panel? Why did you need custom settings?

[MicroframeDev1]

Panels are 1/8 scan. Custom pixel mapping is because the pixels are interlaced: every odd x pixel is 8 y pixels away from every even x one. As I said, this is working. I just finished testing it with a chain of 30 panels.

[MicroframeDev1]

Panels are 64x32, but are apparently wired internally as 128x16. Our previous panels from this supplier were also a virtual 128x16, but had a very different stitching pattern. We switched to the new ones because these are front maintenance.

[board707]

Panels are 64x32, but are apparently wired internally as 128x16.

Yes, this is a completely common situation, and it's described in the library documentation as "quarter-scan" panels.

It's good that you managed to solve this problem yourself, but the library actually has a special example (Pixel_Mapping_Test) for setting up mapping for such panels, which allows you to set up any complex mapping in 2-3 simple steps.

[MicroframeDev1]

Thank you for the tip! I'll be sure to use "Pixel_Mapping_Test" if we have to switch panel types again.

[board707]

Our previous panels from this supplier were also a virtual 128x16, but had a very different stitching pattern.

This is also typical for "quarter-scan" panels—panels with the same parameters, but from different suppliers or batches, can have completely different mappings.

In general, this situation makes publishing specific mappings pointless—and this is why we added an example to the library that allows you to set up almost any mapping from scratch.
The explanation can be found here: #622