STM32L4 Freezes At Freezing Temperatures

tomasf · ‎2019-06-10

Hi Everybody,

I have an issue with STM32L433CCU6, when it's entering or exiting STOP 1 mode at temperatures from -36dC to -29dC uP completely freezes

Basic overview of program:

watchdog is set to 25 seconds
clock: HSE 8MHz -> PLL 16MHz (SYSCLK) -> /2 -> HCLK 8MHz
wake up clock is HSI
Clock Security System is running
uP is going to STOP 1 mode whenever it is possible
uP is woken up by LPTIM external trigger or by RTC every 20 seconds to reset watchdog
BOOT0 pin is grounded
NRST pin is left floating

Overview of behavior:

when STOP 1 mode is disabled uP works from -60 dC to 70 dC without any problem
when STOP 1 mode is enabled:
- from -29 dC to 70 dC: uP works fine
- from -36 dC to -29 dC: after several minutes uP freezes
- from -60 dC to -36 dC: uP occasionally freezes
sysclk is nominal during whole temperature range (USART communication)
VDD is nominal during whole temperature range and also in freeze state
consumption measured at the input of the STEPDOWN converter (from 15 V to 3.3V) is about 4 uA during STOP 1 mode
when uP freezes then consumption rises to value about 100 uA
uP in this mode cannot be reset with NRST
communication with uP through SWD cannot be established using ST-LINK V2 (connected under reset) and ST-LINK utility
when uP is under debug in OpenStm32 environment in moment of freezing GUI shows failure
uP stays in freeze mode also after reaching room temperature (even after several hours)
normal state can be reached only after disconnecting and reconnecting power voltage
after power toggle at mentioned temperature range (from -36 dC to -29 dC), uP runs several minutes and freezes again
this state was observed on 4 pieces of independent PCBs and during several tests in Cryostat or Climatic Chamber

I want to ask whether you experienced similar behavior even at room temperatures, and how to solve this problem.

I will be very grateful if you help me solve this issue. Thanks.

Tomas

tomasf · ‎2019-10-28

Thanks you all for helpful advice.

My problem was hardware related. I found that uP doesn't work even directly with HSE with some crystals and room temperatures. I ran successfully same setup on another design ( same L4 and crystal ) without error, Main cause was insufficient grounding of crystal pads and crystal paths separation from another signals, therefore it was presented on all boards I have tested. It seems that PLL works differently with HSE and can overcome mentioned design error within some temperature range.

Problem was solved using HSI as main clock source as this clock source was sufficiently stable through whole temperature range.

View solution in original post

raptorhal2 · ‎2019-06-10

The chip is rated to -40C. I suggest you discuss with STM if that rating is for run condition when the chip is self heating. (Edited)

Regards, Hal

S.Ma · ‎2019-06-10

Are you using a chamber or a freeze spray? How fast the temperature transient?

tomasf · ‎2019-06-10

Hi Hal,

thank you for joining. Yes that is true, all ICs used in design have -40 dC rating. ST declares that operating conditions are from -40 dC, that should mean it includes all low power modes. I have very good experience with STM32L151CBU6A which works even at temperatures -50 dC without any issue. I performed also cold startup test when STM32L4 was at temperature -40 dC for several hours without power and then power was connected. It performs well. I have tested 4 chips on 4 PCBs all of them have similar behavior.

Tomas

tomasf · ‎2019-06-10

Hi,

I used in first test Cryostat and in several other tests Climatic chamber. Temperature slope in first case was about -1dC / 2 minutes and in second case it was about -1 dC / 1 minute. I always wait for complete settlement of temperature.

Tomas

MikeDB · ‎2019-06-10

Are you sure it isn't the HSE (I assume a crystal) that is freezing ? Maybe stick to the HSI at such low temperatures

tomasf · ‎2019-06-10

Hi,

thank you for your interest. The HSE is monitored with Clock Security System and in the case of HSE failure MSI clock starts and software tries to recover HSE. Even if HSE fails, chip should be able to be reset with NRST, communicate through SWD and consumption should be the same as with HSE. I think HSE works fine; I tested design down to -60 dC without any problem and communication at 1200 bps works, with MSI I think at that temperature it would probably fail.

Tomas

waclawek.jan · ‎2019-06-11

Can't you simply try, sticking to HSI, trying to eliminate one possible point of failure? (Rationale is, what if HSE fails and - as you've said - it's quite likely MSI won't start, so CRS might fail, too).

And I'd also try to omit the PLL. The "big" 'L4s have an erratum where their PLL's won't lock at sub-zero temperatures, if their VCO is set below 96MHz (they are normally allowed to go as low as 64MHz). While in the "little" 'L4s the PLL VCO is not allowed to go below 96MHz, I'd say this is another potential point of failure I'd recommend to try to eliminate.

JW

Viktor POHORELY · ‎2019-06-20

Could you please try to run from HSI to avoid use of HSE ?

Could you also try to run your code from SRAM ?

The problem is sees on 4 boards out of 4 boards, or 4 out of XXXX ?

damh · ‎2019-10-28

Please check your PLL/PLLSAI/... VCO as waclawek.jan said.

Our thread, perhaps same problem: https://community.st.com/s/question/0D50X0000BVnby2SQB/have-anyone-noticed-a-problem-with-the-stm32l4-wakeup-timer-wakeup-from-standby-with-temperatures-10c

The PLL had no problems to lock at our PCBs, but the collateral damage/disturbance seams to be real.