From Per in another thread:
How is the write logic set so that something like this can happen so easily?
It is similar to what I was seeing on a Rev 1 card in my 1000SX before I gave up and sent it back to hargle for a look see. The contents of the flash would change quite regularly...
I haven't re-tested since I rewired the ground to my work area. I had a disconnected ground leg that was causing some power problems.
Hi! I wondered the same thing during early testing on with the prototype but didn't find anything really suspicious. Writes to the Flash memory require a /MEMW in its address region which should not be happening outside of Flash programming sequence. The best I could think of is some sort of spurious write mechanism that is not obvious. It may be poor grounding or bus noise confusing the Flash memory to write when it is not supposed to.
Check out the schematic to see if there is something I missed. I ended up replacing the Flash memory with the 2764 EPROM to eliminate any possibility of memory corruption during early testing. I haven't had any problems since. Some 28C64s support memory protection and the new prototypes also include a write protect jumper.
However, you should be able to reprogram a Flash memory if it has gotten corrupted. If not, there is something else wrong.
Thanks and have a nice day!