Re: In search of bad 4164, 41256 DRAM

On 10/13/19 1:07 AM, Jeffrey Birt wrote:
> Someone sent me some bad 4164 DRAMs which I tried in the DRAM tester this afternoon. This have proven to be very interesting. One of them tested bad on the first set of tests. A second one passed the first set of tests but failed every time there after when the tests were run again. The third one has been tested about 10 times now without fail.
> 
> The testing does not stress the DRAM as far as access time, the number of milliseconds it will retain data with no refresh etc. Actually, the refresh timer runs a little slower than the spec of 2ms but that has not been an issue. There might be temperature or voltage variations that effect the refresh time limit but that is beyond this simple tester.
> 
> Another interesting thing is that filling all cells with 1 or 0 seems to be a pretty poor way to test them. Failures how up much more readily when an alternating pattern is written. Writing the sequence of 0x55 or 0xAA seems to show failures where the filling of 1 or 0 might show no errors.

Yes, if 2 cells next to each other get shorted together for whatever 
reason you will not find them if you fill the whole DRAM with zeros or ones.

But even 0x55 and 0xAA won't find all possible faults. Grab memtest86+ 
and take a look at all the patters it uses (it tells you while it's 
running).

  Gerrit