Curiosity about Data integrity (off topic)

[ortegahernandes]

Good morning everyone.

Most of the protocols available for communication between electronic components on the same PCB (I2C, SPI, etc.) there is no "checksum", that is, what arrived, arrived ....

Some electronic components examples like
eeprom memory is still able to either do redundancy (using two addresses), or do a final check comparing the recorded value and the read value.

In a "Safety" equipment where failures are practically not tolerated, how do these items proceed?

Is it just a charge for good PCB and filtering and noise blocking?

[Ttelmah]

On a well built board, the chance of error should be next to zero. However
at heart that is where programming comes in. For example, I have a set
of parameters stored in EEPROM, that are 'critical'. So the data structure
stored has the data and then a checksum. When the code reads the
EEPROM, it verifies the checksum is correct. If it is not, it tries again. If it
still is not right, it switches back to a 'default' set of values, and raises an alarm.
It also checks that the checksum is right on the data set in RAM. Every
few minutes an integrity check is done, and if this is wrong, it'll then try to
reload the values from the EEPROM, with the same tests applied.
Data transfers to/from a device like EEPROM, should really be as good as
to the processor's RAM itself. However everything has the potential for
error. A particle in particular can flip a cell in EEPROM, in the
processor's own ROM, or the RAM. The physical design of the enclosure
can massively reduce this, and the normal rate is reaching up to 'winning
the lottery' type numbers. You should not get 'transfer' errors, provided the
electrical layout is even remotely competent. If the unit is going into a
plant or lab, where radiation sources exist, this should be considered when
placing the equipment or designing the casing.