ESP32 deep-sleep with LSM6DSOX step-detector: how do you cut false wakes?

[gtrking]

I’m building a small wearable logger: ESP32 sleeps most of the time, wakes on LSM6DSOX step-detector, logs, then goes back to sleep. Looking for settings that work in practice:

• ODR/FS and step-detector config that avoids vibration false wakes

• INT wiring to an RTC-capable GPIO (pull-up value?), latched vs pulsed INT

• FIFO vs direct interrupt; simple debounce in firmware

• Re-init after deep-sleep (soft reset, clear INT, reprogram registers)

• Typical deep-sleep current numbers with the IMU powered vs fully off

Concrete numbers welcome (thresholds, durations, current, wake frequency). Thanks!

[gtrking]

Update: first pass is working better.

• LSM6DSOX: ODR 26 Hz, FS ±4 g, step-detector ON, threshold = 2
  
  
• INT1 → RTC GPIO with 10 kΩ pull-up; added 400 ms cool-down in firmware
  
  
• Deep-sleep sits around ~9–12 µA at 3V3; false wakes dropped a lot
  
  

Hardware reference (external): the board I used is ESP32-WROOM.
If you’ve reached <8 µA with similar wiring, what changed it for you?

[Federica Bossi]

Hi @gtrking,

Lower ODRs reduce power but may miss steps or increase latency. 26Hz could be a good compromise. You can experiment with thresholds 2–4 depending on your environment and motion profile.

FS ±4 g is a balanced choice for wearables, but you could try also ±8g.

Use latched interrupt mode if available (interrupt stays asserted until cleared) to avoid missing events during ESP32 wake-up latency.

In addition, you can disable unused sensor features (e.g., FIFO if not needed).

Ensure sensor is in low-power mode when idle.

Hope this helps!