70% of new boards (STM32L433RCT6) cannot program

[FC_U]

We have new boards with STM32L433RCT6 as MCU and only 30% of boards can program with STM32CubeProgrammer. Other boards just cannot connect to ST-LINK. I have tried these:

1. Use "under reset" mode, not working

2. Use JTAG signals. The good board can be connected and the bad one still cannot.

3. Use UART bootloader with BOOT0 pull up. The good board can be connected and the bad one still cannot. 

I have checked every pin of STM32L433RCT6, and found there is a 2Hz square wave on PB8, I disconnected external connection to PB8, and the square wave still there. If I press reset button, it will go to low. Is this normal?

Can anyone help me? I really appreciate any help. Thank you!

[Tesla DeLorean]

Isn't initial board bring up part of the job?

Review the design again, review the net list, review the power connections. If you have empty boards, or solder-test samples, perhaps probe and check continuity there.

Check part orientation. Check soldering, perhaps X-Ray. Check voltages at power pins.

Insure VDDA/VSSA are powered, check NRST is nominally HIGH

The SWD connection should work unless the pins are non-functional, incorrect, reassigned by running code, or unpowered.

[FC_U]

Thank you for your reply!

I have worked on these boards for over a week, so I have already checked every pin of STM32L433RCT6 and haven't found problem except PB8. 

We have checked SWD signals using oscilloscope, and I have measured every power pins.

Is it normal for PB8 output the 2Hz square wave on a brand new board? Does that mean the STM32L433RCT6 chips are not new?

Could you help to check if the schematic is correct? Thank you!

[Tesla DeLorean]

I'm not aware of PB8 being involved, it is a TIM-CH1 pin, would need to check AN2606 or disassemble the system loader. You could take a working one, erase it, or pull BOOT0=HIGH and see if it also outputs this 2Hz square wave. That would be a quick we to determine normalized behaviour.

By default these don't start an external clock

These were sourced from a legitimate distributor? Any visual difference in markings of ones that work, vs ones that do not? Perhaps a picture, large, clear and in focus, in it's PCB context?

Not sure there's any magic about the LCD models/power. Ive used STM32L432KB/KC in the QFN28 form-factor

[Tesla DeLorean]

https://www.st.com/resource/en/application_note/an2606-stm32-microcontroller-system-memory-boot-mode-stmicroelectronics.pdf

L43x PB8 CAN-RX

Pin designations look good.

Check orientation and shorts

[FC_U]

This is the good one:

This is the bad one:

These two seem same to me.

[FC_U]

Anyway, we will replace STM32 with a new chip bought from Digikey tomorrow. I will post the result tomorrow. Thank you!

[Tesla DeLorean]

Yes, look the same.

Hard to make out the soldering on the left-hand edge of the bad one due to angle/reflection.

I'd tend toward issues with the PCB, or soldering, as all the IC are individually tested for functionality.

Viability tends to pivot around power, and why measuring NRST is top of list.

Other signals on the boot loader pins (see AN2606) may be seen as communication attempts.

Usually exposing UART1 (PA9,PA10) and UART2 (PA2,PA3), as available, can act as secondary proof-of-life test points. ie sending 0x7F pattern at 9600 8E1 and watching for response.

[FC_U]

We have replaced STM32L433RCT6 for all the boards that cannot programming, and now all boards are good!

There's no 2Hz square wave on PB8 for the new boards. Maybe the chips are already programmed?

Is there a way to check whether  the chip is on Level 2 protection?

Thank you!

[Tesla DeLorean · Accepted solution]

I'm not sure of the System Loader behaviour in the Level Two state, but that's where I'd start.