STM32 gotchas
73.Always read the Errata - and not only those directly given for your STM32 model

The Errata form a vital part of the documents stack needed to understand and be able to use a particular STM32 model.

While the Reference Manual deals with the details of the design of particular modules in the mcu, this is mostly a design target description. Inevitably, when implementing this design target, errors occur, resulting in the actual chip behaving differently from the RM's description - and Errata are there to describe these deviations. Sometimes, an erratum is more a documentation clarification rather than error description.

It is not rare to see a developer to struggle with an inexplicable behaviour, just to find out later, that this is consequence of a known deviation from the RM described in the Errata. Therefore, it is a good practice to read the Errata, before starting work on any particular mcu model, and also to follow any changes and updates to those Errata throughout the work.

However, sometimes it pays out to read Errata not only to the given STM32 model, but also to other similar, maybe older models. The reason is, that ST does not update the Errata very timely, especially for "marginal" models which don't see much usage. Even if there is a known problem with an IP module which is present in several models of the same STM32 family and often also in other families, ST does not rush to update all related Errata at once.

Sadly, sometimes it even happens that a known bug of a given IP does not propagate into Errata of a different STM32 family using the same IP at all. For example, in this particular case, an erratum known for the (rather complex and hard to use) I2C module in the 'F1 family did not propagate into errata of 'F4 family using the same I2C IP. According to the 'F1 erratum changelog, the given problem was already known in 2011, the year when 'F4 family was introduced.


Read the relevant ARM Cortex-M errata, too.