New batch STM32F051 LSE don't start

[passafabio2]

A new batch of electronic board mount an STM32F05R8T6

AA170 9U

TWN AA 131 marked.

All of this devices don't start up LSE clock.

The code is the same for about five years and makes the LSE clock start up till this days correctly. The initialization was made using example code bundled with discovery stm32F051.

I create with CubeIDE Version: 1.9.0 a minimal setup (CubeMX) for initializa internal RC but the code hangs on

while(__HAL_RCC_GET_FLAG(RCC_FLAG_LSERDY) == RESET)
 {
 if((HAL_GetTick() - tickstart ) > RCC_LSE_TIMEOUT_VALUE)
 {
 return HAL_TIMEOUT;
 }
 }

The quartz is the same sice five years and i try to modify the capacitor from 15nF to 6,8nF with no success.

Same board with STM32F05R8T6

GQ272 9U

CHN GQ 944 marked

run perfectly.

Any ideas?

Thanks 

Fabio

[waclawek.jan]

> Try replacing the STM32 chip on a new board with one from an old board.

Maybe an interesting exercise would be do the opposite: populate a known good tested working board (such as Nucleo or Disco, or maybe even an older device of yours) with the new 'F051.

JW

[Antoine Odonne]

Hello,

Brief comment to state that 0x2000 is legit code for Rev B or 1. A typo in RM which is under correction, you can refer to product Errata sheet for correct device variants.

Would you please confirm you revision code in addition to marking?

No changes expected around LSE neither.

Thank you and regards,

Antoine

[passafabio2]

STM32F051R8T6

AA170 9U

TWN AA 131

IDCODE 0x20006440

Best Regards

Fabio

[LCE]

So you

a) put an "old" STM32 on a new pcb and it is still not working? Or

b) did you put a new STM32 on an old pcb and it is still not working?

I thought you did a) which would show that the problem is the pcb or the assembly or the LSE crystal.

[passafabio2]

New production batch of board 2022.

Same printed board build at the end of 2021.

Board with "old" STM32 LSE work.

Board with STM32 signed TWN AA 131 munted on same pcb, same quartz, same capacitor, don't start LSE.

[passafabio2]

Today wiring a hybrid 3v3 32Khzxxx oscillator the:

STM32F051R8T6

AA170 9U

TWN AA 131 

IDCODE 0x20006440

Works

So STM thanks for your support and yours fresh detailed documentation about new silicon...

[Antoine Odonne]

Hello,

I am not sure you still await something from our side or the issue is close?

Thank you for clarifying :)

Regards,

[Danish1]

Antoine Odonne

The ST web page for stm32f051 does not show an updated errata document; the current one is ES0202 - Rev 6 - November 2020. So is the "fresh detailed documentation about new silicon" something that is public? If so, where?

Also passafabio reports that the new processors work with an external oscillator - not just an external crystal as they should.

Unless and until both of those are sorted, no I don't think you should close this issue.

[Antoine Odonne]

Hello,

The ES0202 is up to date and describing last silicon on market. No new silicon developed for this line since 2014. Your 2 products are both dated from 2021 and 2019 by decoding your date code. So if you compare both IDCODE I expect you read same value, so same silicon.

The production plant changed as stated in a previous posts, but this is not documented in either RM or ES (besides change of revision on top right corner of package being B,1 or 2) or your packaging location, which neither should lead to LSE being not functional.

We could imagine very small drift in process or electrical charac due to those changes, but it seems impossible that a full batch would be defective.

As suggested by Jan may you check this:

"populate a known good tested working board (such as Nucleo or Disco, or maybe even an older device of yours) with the new 'F051."

In case of defectivity, it should go to (FAR) initiated by your local FAE.

Regarding the LSE itself, in a failed startup state, what voltage do you see on OSC32_IN and OSC32_OUT?

Thank you and regards,

Antoine

[passafabio2]

"populate a known good tested working board (such as Nucleo or Disco, or maybe even an older device of yours) with the new 'F051."

As described before i made this. In a first post I wrote:

Same board with STM32F05R8T6 -> "old device" remains from precedent batch

GQ272 9U

CHN GQ 944 marked

run perfectly.

STM32F05R8T6 has a Discovery board, but the 32Khz768 is not mounted from st.

[Tesla DeLorean]

Right, but I think Jan's point would be to populate parts onto the board. Presumably with parts from ST's BOM, testing the LSE examples, and then perhaps with your crystal.

You look to have gone through a lot of diagnostic steps, if you're confident there's an issue here, you need to push it in via usual support/engineering channels.

Did you initiate a Failure Analysis Report with your local FAE or distributor? I think you're probably seeing a real issue, and that would be the way to advance it, as I don't think you're going to get much further on this thread.

Also pull whatever production/trace codes are on the part reels, and get that to an FAE

[passafabio2]

"Did you initiate a Failure Analysis Report with your local FAE or distributor"...Done by our assembler mfg to their distributor more a month ago. Still waiting...

[waclawek.jan]

My point with "known good" board was that maybe your boards - both old and new - have something which makes them marginal in a way that some parts which came out "slightly better than average" work in both, and the parts which came out "slightly worse but still conforming to specs" don't.

And yes, I meant it as Clive Tesla DeLorean said - use your crystal, and if still no joy, try to use ST's crystal.

Also, post here a minimal but complete compilable code you use for testing, for us to review.

Also, answer Antoine's question.

As both Antoine and Clive said above, at this point you have to talk to ST directly. Note, that we are mere users here, and those who are tagged as ST Employees don't necessarily work as support.

JW

[passafabio2]

The problem seem the poor cleaning made by manufacturer when they find a Sn join between the 32Khz768 quartz pins. Fixing the unwanted join with soldering wire they don't clean well the water based flux and in some occasions made a lower impededence which makes quartz unable to run. The quartz with soldered join is covered with a 16X2 lcd display then detected dirty pin wasn't immediate.
The first batch of stm32f051RBT8 was more strong in oscillating.

Fabio