PDA

View Full Version : S-100 RAM Cards - interpreting RAM patterns



billdeg
January 8th, 2008, 08:32 PM
I have been testing RAM cards lately. I have a good number of RAM cards, and I want to start separating the good from the bad. My test machine is an IMSAI 8080 with a lot of Processor Technologies cards, all working fine and it's a stable system. I use CUTTER to dump memory to the screen and if there's a test program listed in the RAM documentation, I run it. Some RAM cards are a little flakey, but I am happy that most of these cards have aged well.

Here is my question...
When I dump a page, for example 0000 -->0FFF, often I see a pattern something like FF 33 FF 33 FF 33 FF 33 etc.

Is this done by 74L chips on the RAM card just to make it easier to visually observe and differentiate that the RAM is "clear but available" verses FF FF FF FF - depicting that nothing is present in that location/range? In short, what's responsible? Static RAM would not hold memory, so it must be happening when the system initiates.

Are there any mag articles about this topic?

Thanks.

Bill

Dwight Elvey
January 9th, 2008, 11:05 AM
[QUOTE=billdeg;54541]I have been testing RAM cards lately. I have a good number of RAM cards, and I want
---snip---
Here is my question...
When I dump a page, for example 0000 -->0FFF, often I see a pattern something like FF 33 FF 33 FF 33 FF 33 etc.

---snip---

Hi
These are just power up values in the RAMs. They usualy either power up with the same
value in odd and even bytes or for some bits, they are inverted. The inverted ones
are because of the way the RAMs are muxed. The actual cells for alternate addressed
locations are doubly invertered input and output so that the cell has opposite power
up states.
This is most typical in DRAMs but static RAMs may also power up this way.
When testing RAMs, I find it best to remove all but one bank and test that one
all by itself. Failing RAMs often have a tendency to interact with access to
other banks.
Once one knows that all the RAMs in each bank are independently good, test
all banks together. This will test the address decoding on the board.
Most canned RAM test do not test for all the problems that RAMs can have.
March C test is one of the best but it is not much good for retention or problems
with bit line pass transistor leakage.
Still, if I were going to run just one test, I'd run a March C.
Dwight