Why could STM32F103 flash write last so long?

[ROlej.1]

Recently we have noticed in our company problems with writing information into MCU flash. It occured, that write last longer than before and 50 ms watchdog resets device during that. We had to make watchdog delay longer to 100 ms. Now the problem occurs in stm32flash utility. It responds with error in some devices. I'm curious if these cases are related and does anyone knows possible reason? Are there any erratas to STM32F103RBT6? Or are there any known problems with lastly produced MCUs?

[Uwe Bonnes]

Flash Wearout and extreme temperatures could lead to flash times longer than the max value in the datasheet. However page erase time is 40 ms max and that comes near your 50 ms watchdog reset. If the watchdog is driven by LSI , the tolerance of LSI might interfere.

[waclawek.jan]

In addition to what Uwe wrote above, what exactly do you mean by "writing" here? Writing individual words does not take that long, but writing many words and/or erase cumulatively do, so the question is, do you perform watchdog refresh between individual operations?

Note, that there are many counterfeit STM32F103 out there, so if you did not buy them from reputable source, this may be your problem, too.

JW

[ROlej.1]

Write I mean erase + write all the words.

Nevertheless thanks for replies, it might be helpful.