doom_riscv/projects/rgb_panel

RGB Panel driver for iCEBreaker board

This is an example usage of the Hub75 driver IP in this repository and implements driving RGB panels using the iCEBreaker board along with the RGB Panel PMOD.

Default configuration is for a 64x64 panel using 1:32 multiplex. Note that some panels have the Red and Blue channels swapped, so you might have to adapt this ...

This example has 3 modes of operations explained below. Each mode is selected by uncommenting the appropriate define at the top of the top.v file.

Pixel format

The color mapping from the BITDEPTH wide word sent to the core and the data sent to the panels in the various channels is fully configurable by modifying the hub75_colormap module.

The default however is to have the following pixel format for each bit depth:

• BITDEPTH == 8 : RGB332
• BITDEPTH == 16 : RGB565
• BITDEPTH == 24 : RGB888 (meaning with Red channel in the MSB of the word and sent as litte-endian).

Then the channel mapping is to have :

• Channel 0 (hub75_data[3*n+0]) = Blue
• Channel 1 (hub75_data[3*n+1]) = Green
• Channel 2 (hub75_data[3*n+2]) = Red

Check the file data/top-icebreaker.pcf to check that this data channel mapping matches your panels since several pinout have been seen in the wild.

Panel frequency

The build default is to run the panel at 30 MHz. This can be too fast for some panel. During build you can use make PANEL=slow to select PLL settings that will run the panel at 24 MHz instead.

Pattern mode

This generates a Red & Blue gradient across the two axises and then some moving green lines across. Very simple example of generating data directly on the FPGA itself and can also be used as a pretty reliable test that all works well.

Video play mode

In this mode, frames are read from the SPI flash and displayed in sequence.

For this to work, you need some video content to be preloaded into the flash. You can use the special make data-prog target to load a default nyan cat animation.

To load your own animation in flash, checkout the ADDR_BASE and N_FRAMES parameters that tell the module where to look in flash for the image data.

Data needs to be raw frame in RGB565 format, independent of the BITDEPTH parameter. It will internally convert those to the appropriate bitdepth.

Video streaming mode

In this mode, video content is streamed from the host PC to the FPGA using SPI (through the FT2232H used for programming the FPGA).

A control software stream.py is provided in the sw/ sub-directory. It required Python 3.x and pyftdi.

And example usage would be :

./stream.py --fps 10 --loop --input ../data/nyan_glitch_64x64x16.raw

See the --help for other options available.

To prepare content, you can use ffmpeg :

ffmpeg -i input.mp4 -filter_complex "[0:v]crop=540:540,scale=64:64" -pix_fmt rgb565 -f rawvideo output.raw

Obviously the number for the crop filter need to be adjusted for your source material to get a square image that selects the best region to show. Also, you can do use unix FIFOs to directly pipe content from ffmpeg to the stream.py application without the need for intermediate files.

Single PMOD support

By default the code is built to have a double-width PMOD connected on the PMOD1A and PMOD1B port.

There is an alternate single-width PMOD that uses a bit of external logic to reduce the number of IO lines used (so you have more free PMODs slots). To build the project with this option, use make BOARD=icebreaker-single.

The pcf is by default configured to have this PMOD on slot P1A, but you can edit icebreaker-single.pcf to change the pin assignements if it's plugged somewhere else.

If you happen to have two of those PMODs, you can plug them on slot P1A and P1B and use make BOARD=icebreaker-single2x.