Thanks @adamwilt for the pointer here.
I'd like to share my experience in order to possibly help someone else out trying to do this.
I bought a couple of 0.91" 128x32 pixel displays and started down this path using esp-idf and the u8g2 library:
https://github.com/olikraus/u8g2. However couldn't get anything working. The displays would not budge at all - no sparks, no smoke, no signs of life at all.
Well, almost nothing. Just a few clues:
- Getting ESP_FAIL (-1) from calling i2c_master_cmd_begin() in the HAL I2C byte callback.
- With Vcc wired to the OLED, I could not enable flash download. (ie: getting famous: flash read err, 1000). So pretty sure the board is at least drawing power.
- Unplugging the Oled Vcc, causes I2C_master_cmd_begin() to return 0x107 (ESP_TIMEOUT).
Btw... enabling DEBUG level logging was helpful to at least follow the code flow to this point.
First thing I tried to was to disable the internal Esp32 pullup resistors (as these are too week) and add 2 external pullup resistors. Since the Esp32 uses Vcc at 3.3V, I used 2 - 3K3 resistors, one on each of the SDA and SCL lines. The GPIO internal pullup's need to be disabled in the HAL code. (see below).
Btw... using the following GPIO pins:
Code: Select all
#define GPIO_OLED_SDA GPIO_NUM_15 // pin 23 (white-green wire)
#define GPIO_OLED_SCL GPIO_NUM_2 // pin 24 (white-blue wire)
So, after learning how to decode I2C with a 2 channel scope and these 2 excellent protocol resources:
https://en.wikipedia.org/wiki/I%C2%B2C and
https://rheingoldheavy.com/i2c-signals, I found that there was a response Nack on the bus to the first address/write byte written out. Hence the ESP_FAIL.
The scope showed that my code was sending 00111100 from the I2C master (esp32), and reading the 1 bit ack/nack response for a total of 9 bits. Decoding the bits sent on the bus as 7 bit address + 1 bit R/W flag, yields: address = 0011110 (or 0x1E) and R/W = 0 (or write). The ack/nack response bit was: 1 (or Nack).
The reason I rec'd a Nack is due to: "There isn’t anything on the I2C bus that corresponds to the address that was sent. Since no one’s home, there’s nothing to pull the SDA line low, so a NACK is registered." (quoted from 2nd protocol resource above).
So, why is the Oled board not responding?
Well, it seems that the transmitted address is not correct here. I see 0x1E on the bus which is not what I was expecting. I had setup the code to use address 0x3C, as shown here.
Code: Select all
u8x8_SetI2CAddress(&u8g2.u8x8, 0x3C);
Small aside: I found a simple I2C bus scanner here:
https://gist.github.com/herzig/8d4c13d8 ... 6c1306bb12 and used this to determine that the Oled device address is 0x3C.
According to the esp-idf docs here:
https://docs.espressif.com/projects/esp ... s/i2c.html the data argument (ie: the slave address) to i2c_master_write_byte() should be left shifted and OR'ed with the write bit (ie: 0) as follows.
Code: Select all
i2c_master_write_byte(cmd, (ESP_SLAVE_ADDR << 1) | I2C_MASTER_WRITE, ACK_EN);
The hal code was not doing this properly, so the idf I2C driver transmitted the incorrect address (0x3C >> 1, or 0x1E) on the bus. This is why I rec'd a Nack as the first response bit.
Here's the fixed hal code with that will use the proper address as setup via u8x8_SetI2CAddress() shown above.
Code: Select all
/*
* HAL callback function as prescribed by the U8G2 library. This callback is invoked
* to handle I2C communications.
*/
uint8_t u8g2_esp32_i2c_byte_cb(u8x8_t *u8x8, uint8_t msg, uint8_t arg_int, void *arg_ptr) {
ESP_LOGD(TAG, "i2c_cb: Received a msg: %d, arg_int: %d, arg_ptr: %p", msg, arg_int, arg_ptr);
switch(msg) {
case U8X8_MSG_BYTE_SET_DC: //32
if (u8g2_esp32_hal.dc != U8G2_ESP32_HAL_UNDEFINED) {
gpio_set_level(u8g2_esp32_hal.dc, arg_int);
}
break;
case U8X8_MSG_BYTE_INIT: //20
if (u8g2_esp32_hal.sda == U8G2_ESP32_HAL_UNDEFINED ||
u8g2_esp32_hal.scl == U8G2_ESP32_HAL_UNDEFINED) {
break;
}
//i2c_driver_delete(I2C_MASTER_NUM); // delete if already installed
i2c_config_t conf;
memset(&conf, 0, sizeof(i2c_config_t));
conf.mode = I2C_MODE_MASTER;
ESP_LOGI(TAG, "sda_io_num %d", u8g2_esp32_hal.sda);
conf.sda_io_num = u8g2_esp32_hal.sda;
//conf.sda_pullup_en = GPIO_PULLUP_ENABLE;
conf.sda_pullup_en = GPIO_PULLUP_DISABLE; // External pull-up required, 3K3 for Vcc=3.3V
ESP_LOGI(TAG, "scl_io_num %d", u8g2_esp32_hal.scl);
conf.scl_io_num = u8g2_esp32_hal.scl;
//conf.scl_pullup_en = GPIO_PULLUP_ENABLE;
conf.scl_pullup_en = GPIO_PULLUP_DISABLE; // External pull-up required, 3K3 for Vcc=3.3V
ESP_LOGI(TAG, "clk_speed %d", I2C_MASTER_FREQ_HZ);
conf.master.clk_speed = I2C_MASTER_FREQ_HZ;
ESP_LOGI(TAG, "clk_flags 0x%X", conf.clk_flags);
ESP_LOGI(TAG, "i2c_param_config %d", conf.mode);
ESP_ERROR_CHECK(i2c_param_config(I2C_MASTER_NUM, &conf));
ESP_LOGI(TAG, "i2c_driver_install for I2C bus %d", I2C_MASTER_NUM);
ESP_ERROR_CHECK(i2c_driver_install(I2C_MASTER_NUM, conf.mode, I2C_MASTER_RX_BUF_DISABLE, I2C_MASTER_TX_BUF_DISABLE, 0));
ESP_LOGI(TAG, "i2c_driver_install done.");
break;
case U8X8_MSG_BYTE_SEND: { //23
uint8_t* data_ptr = (uint8_t*)arg_ptr;
ESP_LOG_BUFFER_HEXDUMP(TAG, data_ptr, arg_int, ESP_LOG_VERBOSE);
while( arg_int > 0 ) {
ESP_ERROR_CHECK(i2c_master_write_byte(handle_i2c, *data_ptr, ACK_CHECK_EN));
data_ptr++;
arg_int--;
}
}
break;
case U8X8_MSG_BYTE_START_TRANSFER: { //24
uint8_t i2c_address = u8x8_GetI2CAddress(u8x8);
handle_i2c = i2c_cmd_link_create();
ESP_LOGD(TAG, "Start I2C transfer to 0x%02X", i2c_address);
ESP_ERROR_CHECK(i2c_master_start(handle_i2c));
ESP_ERROR_CHECK(i2c_master_write_byte(handle_i2c, ((i2c_address << 1) | I2C_MASTER_WRITE), ACK_CHECK_EN));
}
break;
case U8X8_MSG_BYTE_END_TRANSFER: //25
ESP_LOGD(TAG, "End I2C transfer.");
ESP_ERROR_CHECK(i2c_master_stop(handle_i2c));
ESP_ERROR_CHECK(i2c_master_cmd_begin(I2C_MASTER_NUM, handle_i2c, I2C_TIMEOUT_MS / portTICK_RATE_MS));
i2c_cmd_link_delete(handle_i2c);
break;
}
return 0;
} // u8g2_esp32_i2c_byte_cb
One other thing that took a bit to figure out is: what u8g2 setup function to use for my specific Oled board. There are several options in the library for the 128x32 geometry.
My suggestion is try them all until you find the one that works. Some partially work and display garbage pixels on various portions of the display and some do absolutely nothing. You can see the results I did get in the code comments below.
Code: Select all
/* Configure the u8g2 library and call the setup function for the 128x32 oled module
* U8G2_R0 - no rotation of the display
*/
u8g2_t u8g2;
memset((void*)&u8g2, 0, sizeof(u8g2));
//u8g2_Setup_ssd1306_128x32_univision_1(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // device responds, but no display output
//u8g2_Setup_ssd1306_128x64_noname_f(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // device responds, but no display output
//u8g2_Setup_ssd1305_128x32_noname_1(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // device responds, but no display output
//u8g2_Setup_sh1106_128x32_visionox_1(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // ESP_FAIL, got a nack most likely, did NOT verify with scope
//u8g2_Setup_ssd1306_i2c_128x32_univision_1(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // device responds, address 0x78, get some display output, top line is ok, but garbage after that
//u8g2_Setup_ssd1306_i2c_128x32_winstar_1(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // device responds, get 2 rows of cut off display output, bottom is garbage
//u8g2_Setup_ssd1306_i2c_128x32_univision_2(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // better that 2 above; get more of my text output almost the entire line of "Hello Loop: 0" bottom half is still garbage
//u8g2_Setup_ssd1306_i2c_128x32_winstar_2(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // worse... shift down.. still some garbage... and getting a double line
u8g2_Setup_ssd1306_i2c_128x32_univision_f(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // WORKING: best so far... no garbage, and displays text on the top line, as expected
//u8g2_Setup_ssd1306_i2c_128x32_winstar_f(&u8g2, U8G2_R0, u8g2_esp32_i2c_byte_cb, u8g2_esp32_gpio_and_delay_cb); // this is similar to above, didn't clear the entire display, a bit of garbage at the top, and the line is not in the correct position
/*
* The real address of the device is 0x3C.
* Using external pullup resistors for now... need to do a test with out them -> fails.
*/
u8x8_SetI2CAddress(&u8g2.u8x8, 0x3C);
In summary, there were 4 things I had to do to make this work:
- Disable the internal pullup resistors and use external ones sized correctly for Vcc. (ie: to deliver ~1mA).
- Use the correct I2C address, 0x3C in my case.
- Fix the hal code to properly pass the address to the underlying I2C idf driver.
- Find & test the proper setup function for my ssd1306 compatible Oled.