Stephane and Kai, before I go any further I would like to reiterate the fact that I am still an engineering student and the experience I have is from a relatively short period of time compared to most of the other people here. I don't want to mislead anyone who is making important design decisions for 1 billion products. These are my ideas only.

With that out of the way, I have a few thoughts. I went to Atmel's page and looked at their line of automotive serial EEPROMs. Products like the AT24C01A operate from -55 C to 125 C. Not bad. Stephane, have you looked at ST? They have several R&D and industrial centers in France. Their line looks even better - automotive serial EEPROMs like the M24C01 operate up to 150 C (bare die version, otherwise the top range is 125 C).

I requested retention vs. temperature information from them, but it is from my Yahoo account so they may not take me seriously.

In my designs I always use this multistoring technique to increase redundance. Even when working with the parallel ones. So, I believe that EEPROM will never make trouble BEFORE Flash code memory does.

The approach taken should depend on the probability of failure. While a blanket approach could reduce the chance of bad data, it could be inadequate or simply unnecessary depending on the actual system. I wonder if maybe this is not a serious concern as evidence by the lack of available information?

I would go directly to manufacturer and INSIST on specifications. Tell them, otherwise you will burn down their factory... {grin}

The latter is not a recommended approach. I don't know if that works where Kai is from, but it is a good way to go to prison in the US. You could give ST a call right there in France.

- Lee