Hi Kai,

Statistics is always right.
Maybe the context in witch ( sometimes ) it is used is not properly.
Please observe I don’t said the retention times is 40 years.
I only said “” you can expect “ this life time witch is, as I thing, a statistical approach.
Witch is the probability of this expectation is an other problem.
I only want to make a little commentary .
All serious manufacturers have, generally, two sets of parameters in their data sheets.
Typical parameters ( a statistical approach ) and minimal ( or maximal ) guarantee parameters. Those last parameters are, generally, considered by a good designer.
What this means “” guarantee “ parameters??
Sure is different from “ typical” ones.
I thing the “ guarantee” parameter is a statistical one too but with a probability of occurrence so closed to one that another value then one can not be taken in to consideration.
If not so we could ask the manufactures to pay very serious damages.

As for the problem of hard environments in witch the parameters are guarantee it is also a manufacturer’s bussines.
If the manufacturer guarantee a parameter for military or space applications it is supposed hi tucked in to consideration the extreme temperatures and radiations.
If those sets of parameters are not guarantee but flagged with “ typical” attribute of course the designer must consider other design approaches like ( as you mentioned ) redundant arrays of components ( I thing this method doesn’t transform a “typical” parameter in to a “ guarantee” one )
Generally, when the manufacturer guarantees a parameter this is one or two order of magnitude better then the typical value.
Actually it is viceversa. The manufacturer measures statistical typical parameters and gives us guarantees for values so far from typical that we can make assumptions the probability is one ( 100 %).
If the retention time is guarantee to be 20 years I am sure the typical retention time is ten times better so I can expect ( this time with a probability different from one ) at least 40 years of life ( supposing I respect the data sheet restrictions ).
Of course this is only my opinion based of experience and may be criticized.




Best regards,

Cristian