Rant - Destroying an Arduino EEprom

[udoklein]

At hackaday they describe how someone purposely nuked an EEProm. This time for an Arduino. The last time someone did this for external EEProms. Hackaday as well as both experimenters fail to understand that the collected data is useless. After 1,230,163 cycles one EEProm cell failed to retain data for a short period of time. However the actual failure was most probably ***much*** earlier. The reason is that EEProm failure is characterized by decreased retention time. So what they found is that after 10 times the specified value of write accesses the retention time is ***way*** below the specified retention time. Thus the failure to meet the specification was obviously much earlier.
Of course the experiment did not even consider influences like voltage, temperature or production deviations.

Why is it that such clueless measurements get so much attention?

[AWOL]

Why is it that such clueless measurements get so much attention?

Until you pointed it out, I'd never heard of it.

[udoklein]

I am not an electrical engineer, but Horrowitz Hill as well as some datasheets mention this. So I thought this is pretty much well known. Bascially programming an EEProm many times will increase leakage. Thus retention time gets shorter and shorter. The specified cycles are the guaranteed number of cycles for which the retention time will be as specified. The specified retention time is in the order of years. Thus any experiment that tries to determine when this fails will take a very long time or it has to find a way to either increase leakage in a controlled way or to measure leakage.
Burning an EEprom till the retention time is below the minute range does not tell very much. Except that exceeding the specification will nuke the chip. It does not even proof that the chip met the specification in the first place.

[Coding Badly]

Probably the same reason "Jackass: The Movie" was profitable.  Most of us are essentially clueless monkeys with an insatiable morbid curiosity.

[retrolefty]

I would just call it an example of digital masturbation.

Lefty

[udoklein]

If they want to drive the digital masturbation to higher levels the following questions remain:

How many cycles will the flash memory need to drop below a retention time of 1s?
At which temperature will retention time drop below 1min while NOT destroying the processor? Both question for flash and EEProm.
At which voltage will the processor be destroyed?
At which temperature will the processor be destroyed?
Are there any differences for depending on the pin? E.g. do reset or ACC have a different threshhold?
Which is the minimum voltage / current to explode the processor?
Which is the minimum voltage / current to vaporize it?

And of course: does it blend?

[mowcius]

And of course: does it blend?

Get on and try it - anyone got a decent blender?

[AWOL]

I am not an electrical engineer, but Horrowitz Hill as well as some datasheets mention this

Neither am I, though I was aware of the data retention limitations before.

Why is it that such clueless measurements get so much attention?

It was the clueless measurements that I had never heard of before - thanks for drawing attention to them.   ;-)

[udoklein]

Ah, got the point