latest updates and/or corrections 1-31-06

If you get an "out of memory" error or less than the normal 
number of bytes free (should be 38,911 for the C64) when you turn 
your computer on, that's a RAM problem and usually points to a 
partial failure in one of the RAM IC chips. Blank screen can be a 
RAM problem but can also be caused by many other failures such as 
a bad PLA or power supply which are actually more common. Bad RAM 
can produce a "garbage" screen which is a symptom where the screen 
fills up with random characters when the computer is turned on. 
There are two memory control chips that should also be suspected 
when you have an apparent RAM problem: U13 and U25 in the C64, and
U14 and U15 in a C128 and 128DCR. Those are generic 74LS257 logic 
chips. The C64C integrates all memory control inside its PLA and 
so doesn't use the 74LS257 chips.
     Early C64 computers have eight 4164 TTL RAM chips installed.
Some interim brown case and all later OEM C64C (white case) 
computers have only two 4464 RAM ICs (41464 is the same IC) as 
those chips have four times the memory "density" of the 4164. They 
are not interchangeable with the earlier RAM. Early C128 computers 
have sixteen 4164 chips and a total memory count at startup of 
122365 bytes free in 128 mode. The USA metal case C128DCR has four 
4464 (41464) RAM chips and the same bytes free count on the 128 
mode startup screen. The 64 mode in all C128 computers shows the 
same memory count at startup as a stock C64.
     One common symptom of bad RAM is the error message "OUT OF 
MEMORY IN 0" at startup. That means the computer failed it's RAM 
self test and therefore can't load programs or perform even simple 
tasks. Because it can produce the opening screen, this means one
RAM IC has an open element or stuck bit... not as serious as a 
shorted chip. There are eight RAM ICs in there, so how do you 
find the bad one without just replacing them all one at a time? 
Fortunately, there are several methods.
     It's possible to calculate which RAM bank has failed the self
test using the bytes free number, but I could never find exact 
information regarding that procedure, so I've never used it. Since
the computer partially works (opening screen comes up), I use a 
method that usually works: to "piggyback" a known good 4164 RAM 
chip on top of each computer RAM IC and then power up the computer 
and observe the number of bytes free on the screen. If the value 
changes (even if it's still not correct), change that board RAM IC
and try the computer again. Never just leave the piggyback chip in
place as that may cause timing problems or may eventually stop 
working. The board chip can continue to deteriorate and could 
eventually short out. It's best to remove and replace it now. 
     The piggyback method requires that you install the test chip 
over a board RAM chip so all pins make good contact and no pins 
short together. The test chip must be oriented correctly as well. 
There is a notch or dot (or both) on one end of those ICs. Those
markings must line up because if the test chip is installed
backwards, it will likely be damaged when power is applied. You 
can press the test chip pins against a tabletop to push all pins 
inward slightly, and that will provide some spring tension when 
the test chip is inserted over the board chip and can hold it in
place. It's OK if you want to hold the chip with your fingers and 
then power up the computer. There is only five volts present and 
therefore absolutely no danger of electrical shock. Again, if the 
bytes free number changes, that board IC is probably bad and 
should be changed. If there is no change when all RAM is checked,
suspect the two RAM control logic ICs mentioned above. They must
be replaced to diagnose a problem there.
     There are several ways to replace ICs on a computer board. 
One way that doesn't require a lot of work is to snip all the pins 
off close to the body of the IC with a pair of small cutters and 
just solder the new IC to the remaining pin stubs. It doesn't look 
very neat, but it works fine as long as no pins are shorted 
together. If you get the stubs too hot during soldering, they can 
loosen up in the PC board holes and start moving around... so you 
should solder quickly. Examine your work carefully when you're 
done. Look for shorts between pins and solder splashes or bad 
connections (sometimes called "cold solder" joints).
     The next method is a variation of the first, namely to snip 
all the pins of the bad chip, but in this case to desolder them 
one at a time so you can install a socket. Once the stubs are 
removed, you need to clean the residue of solder from the holes. 
That's easily done with your iron and a solder sucker, a vacuum 
device that slurps up molten solder when the trigger is pressed. 
If a hole is open but a tiny bit of solder remains, you can poke 
a straight pin through and wiggle it around a bit... gently! 
Solder is made of tin & lead and is relatively soft, but if you 
get too aggressive, you may damage a board trace (the copper 
wiring in and around the holes), and that MUST be repaired before 
you install a socket. If you install an IC or socket over an 
unrepaired trace, you've just created and covered up another 
problem that is -very- difficult to find, even for a tech.
     With regards sockets, I always use them when replacing chips.
4164 RAM chips usually cost under two US dollars each so just 
replacing suspected chips by snipping the pins is OK. I always 
desolder larger proprietary (expensive as well as hard to get) 
ICs such as the PLA or SID intact. That way, if the suspected IC
happens to still be good, I have not ruined it by cutting the pins 
off. Removing any chip intact requires a lot more care and the 
proper tools to prevent chip and/or board damage. Once even a small 
IC is removed, it's still a good idea to install a socket. Then 
that area of the board will never be subjected to the heat of 
soldering again. Each board rework with a soldering iron results 
in weakening of board/trace bonds. The less of that the better.
     Note that shorted RAM will always produce a blank screen and
will often get very hot to the touch. RAM normally runs barely warm
to the touch after several minutes of operation. Any that get hot
quickly or even noticeably warmer than the other RAM chips are bad!
One or more shorted RAM chips usually point to a failing power
supply. That MUST be checked before you turn the computer on again
after repairs are done. One easy way is to measure the voltage at
the appropriate pins of the four pin DIN plug of the power supply
while it's plugged in to AC power but disconnected from the
computer. Don't short any pins together while testing!
     Two pins will measure 9 to 11 volts AC, which is normal. The
other two pins should measure 5 volts DC. If that regulated source
is high or low by more than a few tenths of a volt, the supply is
failing and should be replaced. All black or white original CBM 64
"brick" power supplies are potted with epoxy inside and therefore
cannot be repaired. Note that some supply failures will not show up
until the supply has been allowed to warm up for awhile. Therefore
it's best to run the power supply with an artificial "load" (not
a computer!) connected to it such as a small 12v light bulb. See 
the PS tester article for more information about that.
     Other chips may be damaged by overvoltage from a failing C64 
power supply but RAM is particularly sensitive to that problem and 
will usually be the first to fail by shorting out. Obviously the
piggyback method of chip testing will not work with shorted chips
that produce a blank screen. It's interesting to note that shorted 
RAM chips in the C128 are almost unheard of. The 128 switching 
supply is more reliable and rarely harms the computer even when it 
does fail. Voltage "sag" (lower than 5 volts output) when loaded
usually points to failing filter capacitors in the supply. The 
C128 power supply is repairable unlike the 64 epoxy brick. 

Ray Carlsen
Carlsen Electronics... a leader in trailing-edge technology