Please you all out there, have you ever made one thought about the time your memory test may take ?

writing a byte 55, over write it with aa , testing it against aa, .. and that 8192 times ? you must have time in your aplication. and what do you get at the end of the day ? do you get the information about a short between some bits in the adressbus (and or databus) ? NO.

there is a testalg. that is called walking one's and walking zero's. assume that the starting adress is 0001 write a byte to this location, and now let another 1 walk through this adress and read if you get the byte from somewhere else like adress 0003 0005 0009 ... until 8001 ... now the next starting adress is 0002 and you let the 1 walk again through the adress range .... afther this, you do the same with a walking 0. with this you can test against shorts in the adressbus.

also you can test the databus with walking 0's and 1's

the last test is a combined walking 0's and 1's on both adress and databus.

well now you know your hardware around the memory is okay, but you have not tested the whole memory locations. but ALL adresslines and datalines. we test 1MByte with this very fast at each HW-Reset !, and i have never seen that a singel memorylocation inside an RAM is bad, it was always shorts or bad soldered pads on the board that caused the problem.

Search the web for more information about that, as i cant give our source away. but you may look for michael barth, i remember he released some public C sources about that a while ago.

Regards.