@daslolo Yep. And I managed to integrate this to the M5Stack library. I am doing tests, but so far so good. I've run a few modified sketches without issue. The TFT_ArcFill example is quite impressive.

Replacing the Display.h/.cpp files with the TFT_eSPI library was pretty easy. The Display library has a couple of M5-specific functions, but once added to the driver, everything works well:

void setBrightness(uint8_t); //Specific to M5Stack
void sleep();                //Specific to M5Stack

There may be others, I'll see what happens when I compile the non-touch examples.

After a little bit of poking around I managed to turn off the screen:

Wire1.beginTransmission(0x34);
Wire1.write(0x12);
Wire1.write(0b01001011);  // LDO2, aka OLED_VDD, off
Wire1.endTransmission();

This turns off LDO2, and thus the backlight. To turn it back on:

Wire1.beginTransmission(0x34);
Wire1.write(0x12);
Wire1.write(0x4d); // Enable LDO2, aka OLED_VDD
Wire1.endTransmission();

@jimit Now all we need is to have that officially included in the M5Stack library.

Nice list.

Espressif recommends you turn off BT, ADC and WiFi yourself before going to deep sleep, as the ESP32 doesn't do it "gracefully". Something like this:

esp_wifi_stop();
esp_bluedroid_disable();
esp_bluedroid_deinit();
esp_bt_controller_disable();
esp_bt_controller_deinit();
esp_bt_mem_release(ESP_BT_MODE_BTDM);
adc_power_off();
esp_deep_sleep_start();

#define RST_PIN 1 //Pin 1 para el reset del RC522

As @m5stack said, PIN 1 is used by Serial0. You should pick a GPIO that's not in use already.

@jimit That puts the whole machine to sleep. If you want to turn only the LCD off/on, you can do M5.Lcd.setBrightness(brightness); where brightness is 100 for on and 0 for off.

I poked around the APX192 code, and around a datasheet found here. First, there's this code:

void AXP192::begin(void) {
  Wire1.begin(21, 22);
  Wire1.beginTransmission(0x34);
  Wire1.write(0x10);
  Wire1.write(0xff);  //OLED_VPP Enable
  Wire1.endTransmission();
  Wire1.beginTransmission(0x34);
  Wire1.write(0x28);
  Wire1.write(0xff); //Enable LDO2&LDO3, LED&TFT 3.3V
  Wire1.endTransmission();

A peek at the datasheet says that REG 28H is power management for LD02 and LD03. Sure thing. That's what the comment in the code above says – along with a hint that LD03 is for the TFT.

Meanwhile, REG 10H is two ON/OFF switches, for EXTEN and DC-DC2 – commented in the code as VPP_Enable. Bueno.

I am not sure which of EXTEN or DC-DC2 is connected to the OLED (both?), but turning either or both off will cut off the current to the OLED.

Moreover there's this code later:

  Wire1.beginTransmission(0x34);
  Wire1.write(0x12);
  Wire1.write(0x4d); //Enable DC-DC1, OLED_VDD, 5B V_EXT
  Wire1.endTransmission();

And REG 12H seems to do more of the same as 10H:

Bit 3 turns ON/OFF LD03, which, as we saw above, is supposed to be the OLED. Good odds that setting bit 3 of REG 12H to zero would work too.

I don't have time to test it now, but I'll try later tonight.

Very nice. The speaker noise issue forced me to work out a fix to shut up the speaker on demand, without having to recompile every time. Here's how: you add an end() method in Speaker.cpp:

void SPEAKER::end() {
  ledcDetachPin(SPEAKER_PIN);
}

(declared in Speaker.h)

class SPEAKER {
  public:
    SPEAKER(void);
    void begin();
    void end();

Now you can call M5.Speaker.end(); any time to make sure the speaker shuts up. Works with any project – making it a generic solution.

@dda Actually this seems to work better:

  M5.Lcd.writecommand(ILI9341_DISPOFF);
  M5.Lcd.setBrightness(0);

1ILI9341_DISPOFF` doesn't turn off the backlight, just the screen itself. So by doing both, you should be saving more.

While I'm at it, an explanation about M5.Axp.ScreenBreath(x);. The code is:

void AXP192::ScreenBreath(uint8_t brightness) {
  Wire1.beginTransmission(0x34);
  Wire1.write(0x28);
  Wire1.write(((brightness & 0x0f) << 4)); //Enable LDO2&LDO3, LED&TFT 3.3V
  Wire1.endTransmission();
}

What it does is it takes the value you passed, cuts it to 4 bits (0-15) and moves those 4 bits to bits 4-7. Register 28H is voltage control for LDO2 (bits 4-7) and LDO3 (bits 0-3). So the OLED is connected to LDO2.

What that 0-15 value represents is a voltage of 1.8 to 3.3V, in 100mV/steps. So M5.Axp.ScreenBreath(0); sets voltage to 1.8V, and M5.Axp.ScreenBreath(15); to 3.3V. Not what we really want – as the goal was to set voltage to zero... :-)

Whereas register 12H controls the state (ON/OFF) of a bunch of wires, including LDO2 (bit 2). Hence the code above.

@mati I have mentioned this to Jimmi, and his answer basically was, well don't use wifi and OLED. Not exactly a practical answer. I'd rather have a long hat, isolating the DHT and BMP from the heat.

A screenshot or full text of the error would be helpful...

If you have the OLED and WiFi on, the heat they produce will skew the results. I connect the ENV hat to my Stick C with wires, and I get different, lower-temp results.

MOSI/MISO is the SPI equivalent of Tx/Rx for UART (except the lines are not crossed: MOSI connects to MOSI, and MISO to MISO). The other difference is that's it's a bus, like I2C, with one master and one or more slaves. So you can connect a screen and a LoRa chip, for instance. However, you need a separate SS (Slave Select) line per device. You're going to have difficulties finding enough pins for this. As suggested, it'd be much better to use a managed chip (MCU + LoRa chip).

I've tried a few things but it's too cumbersome. I switched back to M5Stack for this.

In my code I was using Wire instead of Wire1... So by changing that it worked.

The datasheet has most of the registers explained. I have the Chinese version, so it requires a little squinting (my Chinese is ok but nowhere near the level required to read datasheets with a smile pasted on my face ;-).

However I recently spent some quality time with the source code and the datasheet, so I have that more or less nailed down.

Register 46H is IRQ Status 3. Bits 0/1 are for the button:

// 0 not press, 0x01 long press, 0x02 press
uint8_t AXP192::GetBtnPress() {
  Wire1.beginTransmission(0x34);
  Wire1.write(0x46);
  Wire1.endTransmission();
  Wire1.requestFrom(0x34, 1);
  uint8_t state = Wire1.read();
  if (state) {
    Wire1.beginTransmission(0x34);
    Wire1.write(0x46);
    Wire1.write(0x03);
    Wire1.endTransmission();
  }
  return state;
}

In Wire1.write(0x03); 0x03 is the two bits 0 and 1 set.

There are many more registers, and the best would be to get a datasheet in English and read the registers description.

It should be 34. From MPU9250.cpp:

// Configure Interrupts and Bypass Enable
// Set interrupt pin active high, push-pull, hold interrupt pin level HIGH until interrupt cleared,
// clear on read of INT_STATUS, and enable I2C_BYPASS_EN so additional chips
// can join the I2C bus and all can be controlled by the Arduino as master
writeByte(MPU9250_ADDRESS, INT_PIN_CFG, 0x22);
writeByte(MPU9250_ADDRESS, INT_ENABLE, 0x01); // Enable data ready (bit 0) interrupt
delay(100);

0x22 = 34.

Nice list.

Espressif recommends you turn off BT, ADC and WiFi yourself before going to deep sleep, as the ESP32 doesn't do it "gracefully". Something like this:

esp_wifi_stop();
esp_bluedroid_disable();
esp_bluedroid_deinit();
esp_bt_controller_disable();
esp_bt_controller_deinit();
esp_bt_mem_release(ESP_BT_MODE_BTDM);
adc_power_off();
esp_deep_sleep_start();

I have a lot of different LoRa chips and setups. I have one board, at 230 MHz, that I can connect through the hat connector (I need 3 pins + Vcc/GND). I have transceivers, both at 230 and 433 MHz, I connect to via RS485 (I bought a hat for that, and also some RS485 for my M5Stack cores). etc etc. The smaller the better :-)

@ws1088 Funny enough I bought my M5StickC with the watch strap and all, but not because I wanted a watch – I don't wear them – but because it enables me to tie it up to my backpack when I do LoRa range tests :-)