ESP32 Forum

Posted: **Mon Jan 16, 2023 4:11 am**

I was interested in how long it takes an ESP32-C3 to enter and wake from light-Sleep, but I could not find anything posted, so i did some experimentation. I thought I would post the results here for others.

I wrote some code to go into a loop entering Light-Sleep and then wake up on a GPIO (4) going high. I also set another GPIO (13) during the process so I could look at the timing on my scope.

Here is the basic loop. It uses the Arduino environment along with some IDF API calls on an ESP32-C3 development board running at 160 MHz.

Code: Select all

  // Go into an infinite loop of Light-Sleeps waiting on a GPIO to wake us up.
  // - GPIO04 - Wake-up Pin - Feed 100 Hz (3.3V) into this
  //          - This is level based, so it will wake up as long as it is high.
  //          - If it is already high, we will wake up right away.
  // - GPIO13 - Sleep-Indicator pin
  uint32_t CpuCountS;     // Count of CPU cycles
  uint32_t CpuCountE;
  uint64_t MicrosS;       // From ESP microseconds timer
  uint64_t MicrosE;
  
  pinMode(GPIO_NUM_4, INPUT);
  pinMode(GPIO_NUM_13, OUTPUT);

  gpio_wakeup_enable(GPIO_NUM_4, GPIO_INTR_HIGH_LEVEL);    // Enable wakeup on high-level on GPIO 4
  esp_sleep_enable_gpio_wakeup();                          // Enable wake-up via GPIO pin level
    
  // Infinite loop
  for (uint16_t i; ; i++) {
      
    while (digitalRead(GPIO_NUM_4) == HIGH) {}             // Loop until the pin is low since we don't want to wake up right away
      
    MicrosS =  esp_timer_get_time();
    CpuCountS = cpu_hal_get_cycle_count();

    GPIO.out_w1ts.out_w1ts = (1 << GPIO_NUM_13);           // Extra-fast - Set HIGH
    esp_light_sleep_start();
    GPIO.out_w1tc.out_w1tc = (1 << GPIO_NUM_13);           // Extra-fast - Set LOW
      
    CpuCountE = cpu_hal_get_cycle_count();
    MicrosE = esp_timer_get_time();
      
    printf("Sleep wake-up via GPIO after: %10.6f secs clock. %10.6f secs CPU.\n",
           (MicrosE - MicrosS)/1000000.0, (CpuCountE - CpuCountS)/(1000000.0 * 160));
  }

  // Cleanup wakeup settings even though this will never be executed in this code.
  esp_sleep_disable_wakeup_source(ESP_SLEEP_WAKEUP_GPIO); 
  gpio_wakeup_disable(GPIO_NUM_4);

I then fed a square wave around 100 Hz into GPIO 4, and watched GPIO 13 to see when the ESP entered and exited sleep mode.

My notes from taking a bunch of measurements are:

Code: Select all

Light Sleep Timing Data

 - Going to sleep:
      +  0 us - GPIO13 goes high - Just before the Sleep call.
      +320 us - GPIO 13 goes Hi-z
              - Voltage drops to Vcc/2 in a couple of us.
              - Drops to zero in < 5us when pulled low by 20K ohms
      +600 us - Drops to zero loaded with a 1x scope probe.
      +900 us - Drops to zero loaded with a 10x scope probe.

  - Basically,  sleep is entered about 320 us after the esp_light_sleep_start() call.

 - Waking up:
     +      0 us - GPIO 4 goes high to wake up the ESP32
     +280-290 us - GPIO 13 goes high again as pin returns to output-high from hi-z sleep mode
     +490-500 us - GPIO 13 goes low, set by code immediately after sleep returns.

 - The time returned by esp_timer_get_time() pretty accurately measures the true elapsed time 
   including the sleep period.
 - The time returned by cpu_hal_get_cycle_count() is always about 250 us over the sleep period
   regardless of the sleep length.
   - That seems to be how long the CPU actually executes during the sleep entry and exit.

So the bottom line is that it takes about 320 us to enter Light Sleep mode and then about 500 us to return to normal execution. there is about 10-20us of jitter in the signals.

Here are a few lines from the Serial console:

Code: Select all

Sleep wake-up via GPIO after:   0.000989 secs clock.   0.000244 secs CPU.
Sleep wake-up via GPIO after:   0.001002 secs clock.   0.000243 secs CPU.
Sleep wake-up via GPIO after:   0.000995 secs clock.   0.000249 secs CPU.
Sleep wake-up via GPIO after:   0.001001 secs clock.   0.000250 secs CPU.
Sleep wake-up via GPIO after:   0.000992 secs clock.   0.000250 secs CPU.
Sleep wake-up via GPIO after:   0.001000 secs clock.   0.000250 secs CPU.
Sleep wake-up via GPIO after:   0.000994 secs clock.   0.000245 secs CPU.
Sleep wake-up via GPIO after:   0.000993 secs clock.   0.000245 secs CPU.
Sleep wake-up via GPIO after:   0.001002 secs clock.   0.000243 secs CPU.
Sleep wake-up via GPIO after:   0.001003 secs clock.   0.000249 secs CPU.
Sleep wake-up via GPIO after:   0.001000 secs clock.   0.000251 secs CPU.
Sleep wake-up via GPIO after:   0.000993 secs clock.   0.000251 secs CPU.
Sleep wake-up via GPIO after:   0.001001 secs clock.   0.000250 secs CPU.
Sleep wake-up via GPIO after:   0.000989 secs clock.   0.000244 secs CPU.
Sleep wake-up via GPIO after:   0.000995 secs clock.   0.000245 secs CPU.
Sleep wake-up via GPIO after:   0.001002 secs clock.   0.000244 secs CPU.
Sleep wake-up via GPIO after:   0.000995 secs clock.   0.000244 secs CPU.

I'll post some scope images in my next post in a few minutes.

Posted: **Tue Jan 17, 2023 5:34 am**

Sorry, it took a while to post the scope images. I'm a new user and my post was sent for moderation.

First, are the traces with a 1.0 millisecond time scale.

Blue is a 1.28 KHz signal being fed into GPIO 4. That is the wake-up trigger.
Yellow is the output from GPIO 13. It is set high just before the sleep and low right after waking up.

What is happening is my code is setting GPIO 13 high, then once the ESP32 actually enters light-sleep GPIO 13 goes high-impedance. Once it goes hi-z it quickly drops to about Vcc/2. Since the only connection to the pin is a 10x scope probe it eventually then drains down to 0.

Once GPIO 4 goes high, the ESP32 starts the wake-up process. GPIO 13 goes high once the ESP32 is driving it high again. Then a while later the code returns from the sleep and sets GPIO 13 low. That is where the wake-up is complete.

Here is the same thing but with a 100us time base so the timing is clearer.

And here it is with a 20k Ohm resistor pulling GPIO 13 low. It goes low immediately after GPIO 13 goes hi-z. You can also see the ~10us of jitter in the sleep and wake-up timing.

Posted: **Mon Aug 21, 2023 6:01 am**

hi,

We have some issues regarding the timing of going to sleep.
while sending the SOC to light sleep the current consumption is still high for another second and only then the current goes down to light sleep mode consumption.

: light sleep, current consumption.; light_sleep.png (23.45 KiB) Viewed 2090 times

testing info:
ESP32-C3-DevKitM-1
the code is simple - stay awake for 350ms then go to sleep, wake up by timer after 5 sec or GPIO button.

have you seen the same behavior? any idea how to cut the time to go to sleep?

thanks

ESP32 Forum

ESP32-C3 Light Sleep Timing

ESP32-C3 Light Sleep Timing

Re: ESP32-C3 Light Sleep Timing

Re: ESP32-C3 Light Sleep Timing