RA8875 chip BUGS!
I discovered several bugs in the chip.
Registers
Register 0x10 (SYSR). Changing bit 3 to 1 should set (accordly datasheet) the 65K color feature.
In real life this set apparently set almost all drawing functions to 65K color BUT drawing single pixel it result in a 256 color!. I spent a lot of time to discover that I need to set bit 3,2 to 1 to solve the problem, sent a note to RAiO to correct datasheet.
Register 0xB3 (should be SSAR3), part of the 32 bit addressing of the DMA start address... Was purposely erased on all last datasheet, still present in many application notes, what the hell I have to do to address 32bit data?
Hardware issues with SPI lines
IF YOU OWN A EASTRISING from BuyDisplay please read this!!
Thanks to the help of The Experimentalist was discovered that those display have pullups on any SPI output! This cause any type of malfunction with other SPI devices and even damage processor pin in some case. The display affected are the 5" and 7", black PCB, see this page
https://github.com/sumotoy/RA8875/wiki/EastRising-and-Buydisplay-SPI-configuration-and-wiring
There's another **hardware issue on MISO** that's a problem only if you are not planning to use any other SPI devices together with RA8875 (example, the SD card holder!), Paul Stoffregen discover the MISO bug that it's not tristate:
https://github.com/sumotoy/RA8875/wiki/Fix-compatibility-with-other-SPI-devices
The chip it's **NOT out-of-range-values tolerant!** (in contrast of the 90% of the other commercial drivers) If a value it's out of range you can experience various screen weirdness like garbage, white screen or chip freeze! This forced me to carefully surround many function with data range checks.
The **Reset Pin** should completely reset RA8875 but in the real world some register it's not reset, the only way it's a power cicle, no matter how much time you pull low the reset pin. Even software reset have the same issue, this it's evident when you switchover libraries (example, this with Adafruit one)
*Memory Clear full should clear all layers memory?* Accordly datasheet seems yes but it doesn't work as it should. On Datasheet, memory clear can clear the current screen or the entire memory but there's very obscure commands like memory start clear and memory stop clear that are not documented... Grrr
During the fix of setRotation issues I found another bug, the RA8875 forget some parameter previously assigned when you touch the scan registers...Again solved by some library workaround
Another bug, the *user uploaded chars cannot be sized*. Once upped if you touch font size you will see garbage, this is clearly another bug, grr.
**The 10Mhz ceiling for some register**
Accordly datasheet the hardware limit for RA8875 should be 12Mhz, I'm currently find out that
The command *DCR_CIRCLE_START has an hardware issue when a SPI over 10Mhz* it's used, the circles result in garbage!
*Registers TPXH,TPYH,TPXYL cannot be writed at more than 10Mhz* or result in garbage!
**drawPixel at 8bit bpp give strange results**
This is a new entry! The RA8875 has internal automatic 16bit to 8bit conversion but only for the setColor! The drawPixel function draw directly a color into RA8875 DRAM but at 8 bit the internal memory it's splitted so injecting a 16bit color data result in a wrong color in some situations. The only (undocumented) workaround I find it's shrink the 16bit data in 8bit and send a single byte as a color information. This is an not well tested workaround so I'm still investigating.
Update: This looks like an hardware limitation and not really a bug, in 8bit mode the RA8875 expect 8bit data and NOT 16 bit, I've included a quick&raw routine for that in library to see if I can workaround this