STM32F3 HSE crystal startup issues

[nlbutts]

We are using an STM32F303VEH6 on a custom design. We are using a ILCX13-FF5F10-16.000MHZ crystal oscillator. We have the PLL and clock gain setup to generate a 72 MHz CPU clock. On some CPUs we have problems where the CPU doesn't appear to try and start up the HSE. We are using the auto-generated STM32CubeMx code generation for our startup code.

Attached is a snapshot from our schematic. We have built about 800 units to date and have at least 10% of them not function correctly.

[waclawek.jan]

There's probably no simple guideline how to tackle this, so just a couple of random thoughts:

• read AN2867, calculate capacitors. Try different capacitors.
• you 've said "doesn't appear to try and start up the HSE" - make that a certainty, device tests to exclude other possibilities. Use the simplest possible test, outputting HSE to MCO, indicating various possible status on LEDs or signals you can measure. Avoid Cube, write your own code.
• on a failing unit, is the problem consistent, i.e. can it be reproduced? No means double trouble, then try to increase the probability of problem. Is the problem related to temperature? Power supply?
• if the failure is consistent, you can try to swap parts (perhaps starting with the crystal) with a "known good" board. Does the problem "follow the crystal"?
• sometimes problems stem from board contamination during manufacturing - try to wash it. Or, contrary, problems might stem from washing, if used
• review board layout. Are the connections as short as possible, including ground returns from the capacitors? Isn't the ground shared with high current/high frequency/sharp edges/pulses circuitry? You might want post relevant portion of the layout

JW