Skip to main content

ESPCLOCK4 - A Simpler Reset Module

In ESPCLOCK3, the following reset circuitry was used:

This generates a short low pulse for the RST pin on the ESP8266 even if the switch is held down, so that the program can read the switch pin after reset. If it it found that the switch is still held down, it can perform a factory reset.

In the ESPCLOCK4, I wanted to try a software-only approach with the ULP. The main idea is the ULP will check the switch (with a little software debounce processing). If it finds the switch depressed, it will wake the main processor, which will then perform a software reset. 

This means we will only need a single pushbutton, with one end connected to an input pin (pulled up), and the other end connected to GND. No need for additional resistors/capacitors.

I connect the reset button to GPIO4/GND, and add the following definitions:

#define RESETBTN_PIN            RTCIO_GPIO4_CHANNEL
#define RESETBTN_PIN_GPIO       GPIO_NUM_4
#define WAKE_RESET_BUTTON       1

The following code is added to init_gpio() to configure the pin as pulled-up input:

void init_gpio() {
  ...
  init_gpio_pin(RESETBTN_PIN_GPIO, RTC_GPIO_MODE_INPUT_OUTPUT, HIGH);
}

Another RTC_SLOW_MEM variable is added to store the wake up reason, which the ULP will use to indicate to the main CPU why it has been woken up.

// Named indices into RTC_SLOW_MEM
enum { 
  ...
  WAKE_REASON, // Reason for waking up main CPU
};

Also add a macro to read the state of a given GPIO pin:

// Read GPIO pin value into R0
#define X_GPIO_GET(pin) \
  I_RD_REG(RTC_GPIO_IN_REG, RTC_GPIO_IN_NEXT_S+pin, RTC_GPIO_IN_NEXT_S+pin)

Now in the ULP, add the following code:

    // Is reset button pressed (active low)?
    X_GPIO_GET(RESETBTN_PIN),
    M_BGE(_RESETBTN_CHECKED, 1),
    I_DELAY(8000), // 1ms debounce
    X_GPIO_GET(RESETBTN_PIN),
    M_BGE(_RESETBTN_CHECKED, 1),
    X_RTC_SAVEI(WAKE_REASON, WAKE_RESET_BUTTON),
  M_LABEL(_WAIT_WAKEUP_RDY),
    I_RD_REG(RTC_CNTL_LOW_POWER_ST_REG, RTC_CNTL_RDY_FOR_WAKEUP_S, RTC_CNTL_RDY_FOR_WAKEUP_S),
    M_BL(_WAIT_WAKEUP_RDY, 1),
    I_WAKE(),
    I_END(),
    I_HALT(),
  M_LABEL(_RESETBTN_CHECKED),
    ...

What the code basically does is to read GPIO4. If it is low, it will wait for a debouncing period of 1ms and read the pin again. If it it still low, it writes WAKE_RESET_BUTTON (1) as the reason code to WAKE_REASON. Then it waits for the RTC_CNTL_RDY_FOR_WAKEUP bit to become 1, which indicates that the main CPU is ready for wake up. At this point, we issue:

- I_WAKE() to wake the main CPU

- I_END() to stop the RTC timer, because we don't want to ULP code to be run again

- I_HALT() to stop the ULP code

In the setup() code, an additional line is added to allow the ULP code to wake the main CPU:

  esp_sleep_enable_ulp_wakeup();
  esp_deep_sleep_start();

The code to handle the wakeup reason is this:

  esp_sleep_wakeup_cause_t cause = esp_sleep_get_wakeup_cause();
  if (cause != ESP_SLEEP_WAKEUP_ULP) {
    // Initial startup
    ...
    // If reset detected, start in config mode
    delay(1000); // allow time for reset button to be released
    pinMode(RESETBTN_PIN_GPIO, INPUT_PULLUP);
    bool reset = !digitalRead(RESETBTN_PIN_GPIO);
    if (reset) { 
       // Reset button is still held down
       // Perform factory reset
    }
    ...
  } else {
    // Wake up due to ULP
    uint16_t reason = RTCVAR(WAKE_REASON); 
    if (reason == WAKE_RESET_BUTTON) {
      esp_restart();
    }
  }

The cause value will tell us whether the main CPU is just starting up, or it is being woken up by the ULP. If it is starting up, it checks GPIO4 to see if the button is still being held down. If so, it can initiate the factory reset logic.

If it is woken up by the ULP, it checks the wake reason to determine the action to take. So if the wake reason is WAKE_RESET_BUTTON, we call esp_restart() to reset the ESP32.

Issue with esp_restart()

After playing around with the code above, it was found that esp_restart() does not always perform a full reset of the ESP32. This resulted in the ULP code not running, or running once and stopping, in a unpredictable manner. This behaviour can be observed every 2 or 3 resets using esp_restart().

I googled around, and this post seems to fit the bill. It also appears esp_restart() has a long history of problems: example, example, example, example. This is because esp_restart() performs a software reset of the unit, which is different from pulling the EN pin low, and sometimes it misses certain things.

After much digging around and hair pulling, I finally found a reliable way to perform a hardware reset of the MCU:

void esp_hardware_restart() {
  rtc_wdt_protect_off();
  rtc_wdt_disable();
  rtc_wdt_set_length_of_reset_signal(RTC_WDT_SYS_RESET_SIG, RTC_WDT_LENGTH_3_2us);
  rtc_wdt_set_stage(RTC_WDT_STAGE0, RTC_WDT_STAGE_ACTION_RESET_RTC);
  rtc_wdt_set_time(RTC_WDT_STAGE0, 500);
  rtc_wdt_enable();
  rtc_wdt_protect_on();
  while(true);
}

This uses the RTC watchdog timer to induce a hardware reset. The idea for this solution is taken from this post and this post

The above code sets the stage 0 timeout for the RTC WDT to 500ms. If the WDT is not "fed" or disabled by the code within that time, a hardware reset signal is initiated. You can see that a hardware reset is performed by connecting the ESP32 to a serial monitor. The same status messages as powering on the ESP32 are displayed when the reset occurs. 

When this solution is adopted, everything works as expected. The ULP code runs without a hitch after each reset.

ESPCLOCK1 / ESPCLOCK2 / ESPCLOCK3 / ESPCLOCK4

Comments

Popular posts from this blog

Adding "Stereo Mixer" to Windows 7 with Conexant sound card

This procedure worked for my laptop (Thinkpad E530) with a Conexant 20671 sound card, but I suspect it will work for other sound cards in the Conexant family. I was playing with CamStudio to do a video capture of a Flash-based cartoon so that I can put it on the WDTV media player and play it on the big screen in the living room for my kids. The video capture worked brilliantly, but to do a sound capture, I needed to do some hacking. Apparently, there was this recording device called "Stereo Mixer" that was pretty standard in the Windows XP days. This allowed you to capture whatever was played to the speaker in all its digital glory. Then under pressure from various organizations on the dark side of the force, Microsoft and soundcard makers starting disabling this wonderful feature from Windows Vista onwards. So after much Googling around, I found out that for most sound cards, the hardware feature is still there, just not enabled on the software side. Unfortunately, to

Hacking a USB-C to slim tip adapter cable to charge the Thinkpad T450s

This hack is inspired by this post . A year ago, I bought an adapter cable for my wife's Thinkpad X1 Carbon (2nd Gen) that allows her to power her laptop with a 60W-capable portable battery (20V x 3A). A USB-C cable goes from the battery into the adapter, which converts it to the slim tip output required by the laptop. Everything works out of the box, so I didn't give much thought about it. Recently, I decided to buy a similar cable for my Thinkpad T450s. I know technically it should work because the T450s can go as low as 45W (20V x 2.25A) in terms of charging (though I have the 65W charger - 20V x 3.25A).  I went with another adapter cable because it was cheaper and also I prefer the single cable design. So imagine my surprise when the cable came and I plugged it into my laptop and it didn't work! The power manager just cycle in and out of charging mode before giving up with an error message saying there is not enough power. After much research and reading the Thinkwiki

Using Google Dashboard to manage your Android device backup

I used to use AppBrain/Fast Web Install to keep track of which apps I have installed on my phone, and to make it easier to reinstall those apps when the phone gets wiped or replaced. But AppBrain had been going down the tubes, and Fast Web Install had always been a hit-and-miss affair. Android's own "backup to the cloud" system had previously been even more unusable. There isn't a place where you can see what has been backed up. And when you setup a new phone with your Google account, you just have to wait and pray that your favorite apps will be restored to the phone. Typically all the stars have to be aligned just right for this to happen. More often than not, after waiting for an hour or so and nothing happens, you just curse under your breath and proceed to install your favorites apps manually via the Play Store. But I just looked again recently and was pleasantly surprised that things are much more civilized now. Firstly there is a place now where you can loo