Write then Read of data eeprom is no proof of saved data

[pfournier]

18f8622 PCWH 3.249

I mistakenly wrote enough times to a section of my data eeprom to wear out 2 bytes.

As a side effect of this I have discovered that if you write to the eeprom then read, the data looks just fine!!!! You only discover there is a problem after you cycle power or reset.

Does anyone know if there a way around this, or is it just the way the silicon is made?
_________________
-Pete

[kender]

[Guest]

You misunderstood my note, and after rereading it I can see how. I don't want the bytes back, I want to know how I can be assured that the data EEPROM really has saved my data without power cylcling or resetting the processor.

[Ttelmah]

The problem is that certain damage modes, can result in an EEPROM, which writes OK, and holds it's data till the power is removed. Then the data is lost, over days, hours, or even minutes. Unfortunately, the write, followed by read test, will see the data as OK, but then latter the contents degrade...
Dealing with this type of failure is hard. One solution, is to add a checksum elsewhere. On a particular application where high reliability was needed, I switched to using a 24:16 Shannon code, writing three bytes for every two that were needed, with a high probability of detecting failure, or recovering the data. Generally, if I remember correctly, failures result in the bits becoming 'set', so something like writing a second bitwise invert of the byte to the next cell, and assuming if there is a difference, when reading the two bytes, and inverting the second, that the one which has the low value stored is the 'correct' result, should work.

Best Wishes