stm32_lse.inc -- small bug/feature Topic is solved

[geoffrey.brown]

I think the following code:

Code: Select all

__STATIC_INLINE void lse_init(void) {

#if STM32_LSE_ENABLED
  /* LSE activation.*/
#if defined(STM32_LSE_BYPASS)
  /* LSE Bypass.*/
  RCC->BDCR |= STM32_LSEDRV | RCC_BDCR_LSEON | RCC_BDCR_LSESYSEN | RCC_BDCR_LSEBYP;
#else
  /* No LSE Bypass.*/
  RCC->BDCR |= STM32_LSEDRV | RCC_BDCR_LSEON | RCC_BDCR_LSESYSEN;
#endif
  while ((RCC->BDCR & (RCC_BDCR_LSESYSRDY | RCC_BDCR_LSERDY)) !=
         (RCC_BDCR_LSESYSRDY | RCC_BDCR_LSERDY)) {
  }
#endif
}

should be:

Code: Select all

__STATIC_INLINE void lse_init(void) {

#if STM32_LSE_ENABLED
  /* LSE activation.*/
#if defined(STM32_LSE_BYPASS)
  /* LSE Bypass.*/
  RCC->BDCR |= STM32_LSEDRV | RCC_BDCR_LSEON | RCC_BDCR_LSESYSEN | RCC_BDCR_LSEBYP;
#else
  /* No LSE Bypass.*/
  RCC->BDCR |= STM32_LSEDRV | RCC_BDCR_LSEON | RCC_BDCR_LSESYSEN;

  while ((RCC->BDCR & (RCC_BDCR_LSESYSRDY | RCC_BDCR_LSERDY)) !=
         (RCC_BDCR_LSESYSRDY | RCC_BDCR_LSERDY)) {
  }
  #endif
#endif
}

In other words, the busy wait should only occur in the case where an external clock is not used. The reason is that the external clock might depend upon hardware that needs to be "brought" up during initialization. It's not clear that the wait is meaningful.

For my application, I copied the (edited) stm32_lse.inc into an include folder that is searched before the system folders, so no harm no foul. But it did take some time to identify the issue when it occurred in 20.0.3 .

[Giovanni]

Hi,

Moving in "bug reports".

Giovanni

[Giovanni]

Hi,

What is the issue exactly? the behavior of LSERDY bit does not depend on LSEBYP, the RM states:

"After the LSEON bit is cleared, LSERDY goes low after 6 external low-speed oscillator clock
cycles."

So it is just a small delay not related to actual oscillator stability, HSERDY does the same.

Giovanni

[geoffrey.brown]

6 cycles at 1khz is a long time. When using standby mode, the whole life cycle is 1) go to standby, 2) wake on event, 2) do something quickly 34go back to standby. So this wait is completely incompatible with minimum power. The tags I build average 200nA when not collecting data and 500nA when collecting data.

Geoffrey

[Giovanni]

Hi,

Change committed, I did the same also for HSE.

Giovanni