home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Phoenix CD 2.0
/
Phoenix_CD.cdr
/
24b
/
mem_help.zip
/
MEMORY2.DOC
< prev
Wrap
Text File
|
1988-05-03
|
4KB
|
76 lines
In the last message on memory, I talked a lot about memory arrays and
how they are arranged. You should go scrounge that mesage up and read
it first, before you dig into this one.
MAKE SURE you take all precautions against static damage. Touch the
unpainted metal insides of the machine for a second or two before you
touch anything electronic. Don't hand a card to someone else- put it
down, have them, touch top of the power supply, and pick it up. Don't
let anything electronic get near regular white styrofoam or white
packing peanuts, or platic bags of any kind. Use "pink poly"
antistatic bags, 3M dark gray antistat bags, or black d-stat bags. Use
common sense, that helps a lot.
I assume you have already figured what column of 9 chips (or 8, if
parity is not used) contains the bad chip. See previous message.
The next problem is of course to figure out which chip in that column
is the baddie. Most computers have a bios that gags on memory errors
but doesn't provide much useful information. It might give you the
address of the failing chip, necessary information, but that's only
half the story. You also need the failing bit number, a value from 0
to 7, or Parity Bit. You might get lucky, and have the power-on self
test die on the bad chip. Then it should tell you the failing bit,
well most POSTs tell you that. Better to get a good memory tester and
let it cook for a few hours. There is a real good one floating around
as shareware called RAMTEST3 and if you can find it, use it. It tests
EMS ram, conventional, expanded too. If you can't get it anywhere
else, join the EXEC PC BBS at 414-964-5160. If I uploaded this message
to your BBS, chances are I uploaded RAMTEST3 as well- look around.
Anyhow, a ramtest should give the bit number of the failure. What I
often do if the bit number is uncertain or if it seems to be the
parity bit is to shift the entire column up one chip and see if I've
identified the right column. The error should move to another bit if
so. If the error bit is erratic or moves around between tests, then
it's likely the parity chip is bad. Or several chips are quite seated
snugly in their sockets. Try pushing them all down into the sockets
firmly.
If your memory test gives a bit number, that should be the chip within
the row thats bad. Often, they are numbered 0 1 2 3 4 5 6 7 P, so the
first one is 0, etc. Which end is which is anybody's guess; memory
board makers can lay out a board any way they like. Look for hints on
the board itself, sometimes the bit numbers are silkscreened on the
board in tiny white letters, or green ones.
If your memory test gives the bad bits as a bit pattern it will look
something like this:
0101 1010 WRITTEN
0101 0010 READ
0000 1000 XOR although many don't give this line. As you can see,
7654 3210 (bit 3) the bit that's different is #3. Replace the fourth
chip in the column and test again. If a different
bit is implicated each time, it might be the parity chip, or maybe the
address decode or refresh logic on the board has gone south. These
chips are almost certainly soldered in, so buy a new board or get a
good scope and and Electrical Engineering degree.
Things you should try before anything else: tamp all the memory chips
down into the sockets. Don't flex the board to much-- it will break.
Or a marginal trace will let go and you will have a real flakey board
on your hands.
Take all the boards out of the computer one at a time and check the
gold fingers at the bottom. Any that look real grungy should be attack
with a clean pencil eraser. Don't touch the fingers-- skin oils make
them oxidize. Sometimes just pulling the card and looking hard at it
will fix it. Really.
Good luck,
Thomas M. Peters
Stevenson's Office Equipment
4517 N. Oakland Ave
Milwaukee, WI 53211
414-964-6330