STM32L4 LPTIM as pulse counter of intermittent pulses fails when initialising with no pulse

[WSart.1]

I'm working with LPTIM as a Pulse Counter on a STM32L432KC, I have a rain gauge connected to LPTIM1 (PB5) and a anemometer to LPTIM2 (PB1). Both configured on STM32CubeMX as Standalone: counts external clock events and pins assigned as LPTIM?_IN1 using LSE clock.

To simplify, let's just use the anemometer but the symptoms is happening on the rain gauge as well.

Initializing the counter:

HAL_LPTIM_Counter_Start(&hlptim2, 0xffff);

This returns HAL_TIMEOUT when there's no pulse (no wind). If I turn the anemometer with my hand or is winding, when the device is initializing, it returns HAL_OK and it works just fine after that.

Troubleshooting, I found where it's timing out. On stm32l4xx_hal_lptim.c,, HAL_LPTIM_Counter_Start function, you can find this code:

  /* Load the period value in the autoreload register */
  __HAL_LPTIM_AUTORELOAD_SET(hlptim, Period);
 
  /* Wait for the completion of the write operation to the LPTIM_ARR register */
  if (LPTIM_WaitForFlag(hlptim, LPTIM_FLAG_ARROK) == HAL_TIMEOUT)
  {
    return HAL_TIMEOUT;
  }

There's application note AN4865 related to use LPTIM as a pulse counter and it doesn't say anything about LPTIM_ARR register neither LPTIM_ISR waiting for ARROK flag (4th bit). Also, on the example code, they expect a constant pulse signal arriving so this problem never happens.

If I comment if (LPTIM_WaitForFlag... block, it works perfectly.

I opened the issue on github: https://github.com/STMicroelectronics/STM32CubeL4/issues/24

[TDK]

The reference manual explains the logic happening here.

 

Code seems to be doing exactly what it should be doing, per the RM.

[WSart.1]

To what I understand than, the if block commented can cause "unpredictable results".

Would you have a recommendation to my use case? I'm not by far an electronics specialist, it's just a hobby for me, and I ran out of ideas.

I thought on an extra pin attached to the anemometer pin, I bit-bang some pulses (or get another timer to generate a slow PWM), starts the LPTIM, stop the pulses after the successful start and I would be good to go. But I think this is a very ugly hack.