Support EPD7in5_V2

[antonatosn]

Hi,

Please could you also support the new panels 7.5 V2 Black / White
https://www.waveshare.com/7.5inch-e-Paper-HAT.htm

Thanks

[ugomeda]

Yes, but in the future.

I'm planning on abstracting the EPD to support more displays. I have a EPD2in7 i'll implement first, then i'll ring you up to test your display if you're okay with that.

[antonatosn]

Do you know the differences on the driver between v1 and v2. May I can help in development

[ugomeda]

The changes are quite small, it shouldn't be that difficult. Take a look at their repos here :

• https://github.com/waveshare/e-Paper/tree/master/Arduino/epd7in5_V2 (your version)
• https://github.com/waveshare/e-Paper/tree/master/Arduino/epd7in5 (the current version)

I took their code and change it a bit to put it in https://github.com/ugomeda/esp32-epaper-display/blob/master/epaper-esp32/src/epd.c

• EPD_7in5__init() is int Epd::Init(void)
• EPD_WaitUntilIdle is void Epd::WaitUntilIdle(void)
• etc

The pins are the same, you should only have to change the commands sent to the display.

Example :

  EPD_SendCommand(0x01, "\x37\x00", 2);     //POWER_SETTING
  EPD_SendCommand(0x00, "\xCF\x08", 2);     //PANEL_SETTING
  EPD_SendCommand(0x06, "\xC7\xCC\x28", 3); //BOOSTER_SOFT_START
  EPD_SendCommand(0x4, NULL, 0);            //POWER_ON
  EPD_WaitUntilIdle();

Should be replaced by

  EPD_SendCommand(0x01, "\x07\x07\x3f\x3f", 2);
  EPD_SendCommand(0x04, NULL, 0);
  EPD_WaitUntilIdle();

Which is a translation of this code in Waveshare's implementation

    SendCommand(0x01); 
    SendData(0x07);
    SendData(0x07);
    SendData(0x3f);
    SendData(0x3f);

    SendCommand(0x04);
    DelayMs(100);
    WaitUntilIdle();

If you get it working, please share the code :)

[antonatosn]

I have change the resolution to 800x480 on both server and ESP. but now the lodepng_decode throws an error 83 memory allocation failed
on server also I have convert the image to grey-scale for test

[ugomeda]

I already had this problem, waveshare's ESP32 doesn't seem to have the 4MB additional memory and it fails. I developped a workaround at some point by loading the image into chunks, but it was very hacky. I fixed it by using a proper 2-bit image.

Loading the 2 colors 800x480 image in-memory only takes 2x800x480 bits, 94kB. This should fit easily into the 512 kB RAM of the ESP32. I'll try to find where i'm wasting so much memory tomorrow.

[ugomeda]

In the meantime, if you want to implement the color display, you can send a smaller image, ie a 800x240px image. It should display the top of the image, and random garbage after.

[antonatosn]

Ok, I manage to convert it to B/W and use 1 bit of depth. and It worked.
But when I tried to send the image to the screen, esp got kernel panic

0;32mI (2734) EPD: Received new PNG of size 6654␛[0m
␛[0;32mI (2754) EPD: Sending image␛[0m
Guru Meditation Error: Core 1 panic'ed (LoadStoreError). Exception was unhandled.
Core 1 register dump:
PC : 0x400d1fc4 PS : 0x00060630 A0 : 0x800d2110 A1 : 0x3ffc78f0
A2 : 0x3ffe436c A3 : 0x00000320 A4 : 0x000001e0 A5 : 0x0001bc94
A6 : 0x00000001 A7 : 0x00000005 A8 : 0x40000000 A9 : 0x3ffc7917
A10 : 0x00000000 A11 : 0x00000028 A12 : 0x00000027 A13 : 0x00000000
A14 : 0x00000000 A15 : 0x00000000 SAR : 0x00000011 EXCCAUSE: 0x00000003
EXCVADDR: 0x40000000 LBEG : 0x400014fd LEND : 0x4000150d LCOUNT : 0xfffffffe

ELF file SHA256: cbcd4695c23db55df46f0af9d15a600fabee5c44e25d967637d9e0ca1451bfa8

Backtrace: 0x400d1fc1:0x3ffc78f0 0x400d210d:0x3ffc7950 0x400d1b1d:0x3ffc7a00 0x400d1bcc:0x3ffc7aa0 0x400904d9:0x3ffc7ae0

Rebooting...
ets Jun 8 2016 00:22:57

[ugomeda]

Looks like you're trying to access memory which was not allocated. Probably because the PNG was loaded using 1-bit depth, but you're trying to read 2-bit depth images in the EPD code.

In epd.c :

  // 2-bit image, each byte contains 4 pixels
  for (uint32_t i = 0; i < width * height / 4; i++)
  {
    ...

Adapt the code after to read the proper indexes in the buffer.

[antonatosn]

could you explain a bit how you map the image ? It looks a bit complicate to me the EPD_loadImage

[ugomeda]

Yes, it's not straightforward since i have to translate from a 2-bit image to some weird 4-bit image. There is also a buffer to avoid sending the data pixel by pixel.

I'll try to explain as i write the code examples.

So, you're trying to draw a 1-bit image (= 8 pixels per byte) onto your 4-bit display. You'll have to go through the array (width x height) / (8 * 1-bit) times :

  for (uint32_t i = 0; i < width * height / 8; i++)
  {

Then, extract each bit ((image[i] & (0x01 << X)) >> X) and map it to the value the EPD expects (PIX_MAPPING[...]) :

    uint8_t pix_1 = PIX_MAPPING[(image[i] & (0x01 << 7)) >> 7];
    uint8_t pix_2 = PIX_MAPPING[(image[i] & (0x01 << 6)) >> 6];
    uint8_t pix_3 = PIX_MAPPING[(image[i] & (0x01 << 5)) >> 5];
    uint8_t pix_4 = PIX_MAPPING[(image[i] & (0x01 << 4)) >> 4];
    uint8_t pix_5 = PIX_MAPPING[(image[i] & (0x01 << 3)) >> 3];
    uint8_t pix_6 = PIX_MAPPING[(image[i] & (0x01 << 2)) >> 2];
    uint8_t pix_7 = PIX_MAPPING[(image[i] & (0x01 << 1)) >> 1];
    uint8_t pix_8 = PIX_MAPPING[(image[i] & (0x03 << 0)) >> 0];

Looking at https://github.com/waveshare/e-Paper/blob/master/Arduino/epd7in5_V2/imagedata.cpp, it seems like the values are 0x00 and 0x03 for black/white, so no need to change the WHITE, BLACK or PIX_MAPPING constants.

You then have to put the new 4-bit values into the buffer :

    buffer[buffer_pos++] = pix_1 << 4 | pix_2;
    buffer[buffer_pos++] = pix_3 << 4 | pix_4;
    buffer[buffer_pos++] = pix_5 << 4 | pix_6;
    buffer[buffer_pos++] = pix_7 << 4 | pix_8;

So this might be able to load the image into the display memory :

void EPD_loadImage(const uint8_t *image, const unsigned int width, const unsigned int height)
{
  char buffer[BUFFER_SIZE];
  int buffer_pos = 0;

  // 2-bit image, each byte contains 8 pixels
  for (uint32_t i = 0; i < width * height / 8; i++)
  {
    // each byte = 2 pixels
    uint8_t pix_1 = PIX_MAPPING[(image[i] & (0x01 << 7)) >> 7];
    uint8_t pix_2 = PIX_MAPPING[(image[i] & (0x01 << 6)) >> 6];
    uint8_t pix_3 = PIX_MAPPING[(image[i] & (0x01 << 5)) >> 5];
    uint8_t pix_4 = PIX_MAPPING[(image[i] & (0x01 << 4)) >> 4];
    uint8_t pix_5 = PIX_MAPPING[(image[i] & (0x01 << 3)) >> 3];
    uint8_t pix_6 = PIX_MAPPING[(image[i] & (0x01 << 2)) >> 2];
    uint8_t pix_7 = PIX_MAPPING[(image[i] & (0x01 << 1)) >> 1];
    uint8_t pix_8 = PIX_MAPPING[(image[i] & (0x03 << 0)) >> 0];

    buffer[buffer_pos++] = pix_1 << 4 | pix_2;
    buffer[buffer_pos++] = pix_3 << 4 | pix_4;
    buffer[buffer_pos++] = pix_5 << 4 | pix_6;
    buffer[buffer_pos++] = pix_7 << 4 | pix_8;

    if (buffer_pos == BUFFER_SIZE || i == width * height / 8)
    {
      EPD_SendData(buffer, buffer_pos);
      buffer_pos = 0;
    }
  }
}

I hope it helps !

[antonatosn]

why in the last pixel you use 0x03 and on the other 0x01?
uint8_t pix_7 = PIX_MAPPING[(image[i] & (0x01 << 1)) >> 1];
uint8_t pix_8 = PIX_MAPPING[(image[i] & (0x03 << 0)) >> 0];

also as i understand you get 1 pixel / bit from image (so it should be black or white (0x00 or 0xFF) and then what do you try to do???

[antonatosn]

also I have done many trials with the code. I dont have any errors or kernel panic but it shows just random pixels

[ugomeda]

Can you fork the project and push your modifications so i can have a look ?