Raghunathan wrote:

That curve on data retention was VERY interesting.

Well, I published plot of 'data retention versus temperature' curve of ST6220C, because I wanted to show, what the statement '10 years data retention time, guaranteed' can mean!
Specifications about data retention time are extremely different, when having a look at those many datasheets arround. And it's always wise to analyze very carefully what actual definition of this specification is.

Happily there are many modern microcontrollers arround, which show very much better data retention specifications, and which do not make trouble in a 'california desert application'.


Manfred wrote:

this curve shows the retention time for an EPROM.
It is the same curve for FLASH-memory?
In the datasheet AT89S52 is written: Operating temp. is max 125°C, that means the
retention-time is less than half a year -- is this correct, this IC is not good for my purpose: Operating temperature of my unit 24 hours every day max. 85°C: ca. 2 years retention-time - no sufficiente.
And if this curve is correct I do not believe, that retention-time is statistical founded.

In datasheet it's stated, that curve was fabricated from statistical analysis!
Activation energy used by Arrhenius equation is the same for FLASH memory and EPROM, namely 0.7eV. So, Flash memory would show the same decay of curve. But for AT89S52 whole curve is shifted towards much higher data retention times! Data retention time of AT89S52 is much much better than ST6220C!


Cristian wrote:

Also, I think, it is no meaningless to suppose the “Typ” value we can expect ( statistically ) is much higher then 10 years ( the “Typ” field is no filed in your table but it is a common sense to believe it is higher then “Min”)
I thing also the circuit doesn’t function all its life at the extreme conditions .
If you have another opinion please let me know as I am also very concerned of this problem ; not of data retention time but the endurance of some 8052 derivatives.


Actual data retention time for your application can be calculated very easily, when reading the following two papers:

1. Reliability report of actual used microcontroller, and
2. http://www.dmea.osd.mil/ssb1_paper.pdf, where you will find some relevant statistical formulas.

Endurance of FLASH program memory of microcontroller normally is much smaller than of regular EEPROM chip. If endurance is specified to be 1000 times for a microcontroller, this is quite acceptable: Who is changing code as often as 100 000 times?
But for regular EEPROM chip endurance of 1000 times would be much to small. There are many applications, which need much higher endurances...
So, relatively small endurance for FLASH code memory of microcontroller makes sense and results in smaller chip area and reduced costs.

Kai