it might not be the program failing.

it could be bad power, or a mechanical problem with the board.

once i filled spare program locations with little traps (JMP *), and a h/w engineer hooked up a logic analyzer to the 8052 address bus, to see where our spontaneous reset problem was, in the code, or with the hardware. while we were watching, it hit a trap. then it jumped to another trap! obviously not the code. another engineer had bumped the workbench while we watched the analyzer. the problem was traced to bad 8052 daughterboard connections.

but all kinds of unexpected things could happen, in the code or in the hardware, or in what the code expects of the hardware. so, to be safer, you should reset external RAM as well (but then, poof!, wdog_counter ...)

i think you could put the counter in flash eeprom. outside of where you have parameters stored, to avoid failing checksums of those parameters.

there are a LOT of discussions about wdogs on the forum:

Search for: WATCHDOG returned 718 pages

search 8052.com forum for watchdog


james