STM32 chips marking contains a datecode which identifies the year and week when given chip was manufactured. Contrary to the 4-digit YYWW format used by many chip manufacturers, ST on STM32 uses a 3-digit format, YWW, where ony the last digit of the year is used.
STM32 are now around for more than a decade - and given ST's 10-year commitment it is likely that some STM32 models will see more than 20 years of being manufactured. And that together with single-digit year means, that some datecodes will inevitably repeat. This makes identification of the exact manufacturing date a challenge.
This of course is not a problem with the "usual" (and undoubtedly preferred) business model, where a relatively large batch of chips is bought directly from ST or through some of the primary bulk distributors and immediately put into production. Problems arise when "old new stock" is used e.g. as it happened not uncommonly during the "Great Parts Shortage" of early 2020s; or when older equipment containing STM32s is to be maintained or refurbished/updated/reprogrammed, which will hopefully occur at increasing rate in the future, as that's how real ecology works.
Examples of cases when the manufacturing date may be relevant include:
- STM32F303xB/C erratum Imprecise VREFINT calibration values, which pertains to datecodes less than 628, meaning chips manufactured before week 28 in 2016.
- In 'F105/107, according to AN2606, bootloader on devices whose date code is lower than 937 is in version 1.0, which has severe flaws; in this case this is 2009 week 37.
Recently, ST started to add QR-codes to the marking, and this - together with other parts of the marking - supposedly allows to identify exactly the manufacturing time. However, structure of this marks and their exact meaning is not documented publicly, so such identification is possible only through requests through ST.