Why do MRV11-D's crap out?

A little clarification of Dwight's comment on EPROM life expectancy: The 10-year spec is for the half-life of the charge on the floating gate of the EPROM. In other words, after 10 years, you can expect half a programmed gate's charge to have dissipated. This, by no means, renders the data unreadable: a smaller fraction of the charge is sufficient for the EPROM circuitry to read a programmed cell (most algorithms do a 2x to 4x overprogrammming once a cell reads back programmed). The programmed/unprogrammed read threshold is also temperature sensitive, so cooling off a warm part may allow data recovery from an EPROM that's fading away. Also, bear in mind the specifications are for worst case: the charge leaks away fastest as the device sits near its maximum rated temperature (people have erased EPROMs by heating them over a light bulb!). The cooler the chip, the less leakage. The relationship between temperature and leakage is exponential, too: a few degrees make a lot of difference. If Ed's 2764As were running hot, the gates might have lost enough charge in 15-20 years to forget a few bits... This helps explain why all the EPROMs I programmed in 1985 are still intact. They're backed up on floppy... which decays faster? Bob Dwight> Most EPROMs are rated for 10 years. That doesn't mean that Dwight> every one will last that long. Also, many of these types Dwight> of systems were all made with EPROMs from the same lot. Dwight> If there was a long term issue, they'd all fail.