It seems that every 2 to 3 months I have to replace the CMOS battery
in this computer. They used to last years. This is an IBM P3 1000MHZ
motherboard from around 2002. Otherwise it works well.
I read on some discussion group that people should not use a power
strip to shut off their computers because this drains those batteries.
I DO use a power strip to both save electricity and moreso because I
am in a rural area prone to lightning strikes.
I guess I'd rather pay $10 a year for batteries than risk lightning
and waste power too, but I still think they should last longer. I am
buying Energizer batteries too, not some cheap junk. They are CR2032
(3 volt) and are the correct battery.
This is my second IBM motherboard, and the boards are well made and
last a long time, but they both ate batteries, although this one did
not do this the first few years I had it.
Any ideas or suggestions?
Thanks
In my experiences here, with a small group of computer (most of which
are disconnected from power all the time), a CR2032 lasts 3 years.
And that is with different brands. The 3 years is an average. I don't keep
careful track, so that's just a guess.
2 or 3 months implies something is wrong.
A P3 1000MHz would be around the 440BX chipset era. The PIIX4E chipset
spec (Southbridge, uses RTC battery), lists these specifications.
29054801.pdf
Vcc(RTC) Battery Voltage Min 2.0V Max 3.6V
Icc(RTC) Battery Standby Current Min 6uA Max 8uA @ 3.0V
A CR2032 has a capacity of 220mAh (milliamp hours). The RTC draws 0.008 milliamps.
220 / 0.008 = 27500 hours. Divide that by 24 hours per day and get 1145 days.
Divide that by 365 (storing the computer in a closet with no AC power)
gives 3.1 years. So that is the amount of time you should be getting.
It means there must be a "leak" on the VccRTC rail.
Now, if you download this, and go to page 18, you can see what electrical
loads are connected to the battery. JP17 would normally have a jumper from
pin 1 to pin 2, to connect the coin cell to (L16) VBAT. That would be the
6 to 8 uA (microamp) load. But further to the right, they're running a
74HC112 off that battery as well. That opens up a whole world of possibilities
for leakage, such as backfeed into other adjacent circuits.
http://www.intel.com/design/chipsets/designex/BXDPDG10.PDF
To give a grand example of backfeed at where I worked, we designed a system
which had many kinds of logic chips inside it. Some of the logic was low
speed, so CMOS was used.
One day, the project manager turned off the power on one subsystem (he loved
to invent tests for the demo system and make it fall on its face). Now,
the power was half-off, but one of the system modules status LEDs were still
lit up and the circuit was still functioning. What was happening, was
all the logic signals between modules were "leaking" and passing current
to the other (unpowered module). It managed to suck enough power, to make
a 3.6V rail for itself, on a nominally 5V circuit. So in other words, you
kill the 5V supply it normally receives, and like a leech, it managed
to suck enough juice from another circuit, to make 3.6V for itself. Most
amusing. And that's the problem with the 74HC112 on page 18 - it would
have to be carefully tested, to make sure it doesn't create a battery
leakage path. The 3.1 year battery rating is only valid, if VccRTC is
the only load. The 74HC112, depending on what happens to it over time,
could change that radically.
More modern motherboards, don't have outboard logic chips like the 74HC112,
so there are fewer leakage paths. The RTC inside the Southbridge, is
actually cleverly designed, to prevent leakage paths (interfaces
use transmission gates to cut the path).
Even if you measured the current flow coming from the battery socket,
it doesn't guarantee you'd be able to figure out where the current
is going. It would be damn hard to trace.
A possible solution, is to use a battery with a much larger milliamp-hour
rating. Good luck with that
Some electronic devices, years ago,
used nickel cadmium batteries for this purpose - the circuit was
supposed to charge up when the equipment was used, and the nicads
would keep settings powered while the line power was removed.
Invariably, the nicads would leak, the circuit board would be
corroded and ruined. So you can see, other battery types as
a solution, aren't always the best. The user is not told there is
a nickel cadmium battery inside, until there is damage and it
is too late. The modern CR2032, by comparison, is a relatively
clean solution.
Paul
