AD590S / AD590 Long Term drift

Q

Why is the long term drift in the Mil-Spec of the AD590S given as +/-0.1degC
per month but in the datasheet as absolute value?
For which lifetime is the long term drift given in the datasheet applicable?
How can the temperature drift be calculated for a lifetime of 15 years?

A

The AD590 datasheet guarantees that the error due to long term drift will be
+/-0.1degC max. This assumes a constant +5V supply and a temperature up to
+125degC. Long term drift is not generally linear with time. Instead it follows
a semi random pattern meaning it will drift up a little then down a little. On
average the drift is proportional to square root time. The AD590 max drift spec
of +/-0.1degC means the drift will be less than 0.1degC over 5 years over a
temperature range –10degC to +125degC. If the range is extended up to +150deg
the drift may increase.

To get some typical numbers on this for 15 years would require an analysis of
the HTOL data over 1000 hours and extrapolating what this would be using some
statistical techniques to come up with a number for 15 years. However, this is
only a limited sample so could only be used as another reference input to the
above estimation based on the specification.