VanguardLH said:
Pulling the power cord is not required (to remove the 5-volt standby line
from an ATX-style PSU used for the power-on and wake-on-<event> logic on the
motherboard). You are not shorting a voltage line. You are dropping a
signal pin to ground or dropping the voltage across the logic to let the
memory drain; however, there is limiting resistor to prevent shorting the
power or battery to ground.
Although I don't have a circuit diagram for reference, why would the battery
exist if there were power across CMOS when the host was powered off (but
still connected with the cord)? The battery is still needed when you
soft-power off the host which means the standby power from the PSU for the
power-on logic is not available to the CMOS/RTC circuit.
I recommend pulling the power cord, as a simple way to ensure that the
computer is completely un-powered. When dealing with people on the Internet,
you can't be sure they'll follow all the instructions well. Rather than
ruin their motherboard, and have them complain later, it is easier to
just get them to pull the power cord. It is an easy rule to remember.
I dealt with one person, where it turned out his power switch had failed
in the ON position. That has only happened the once, in the time I've been
helping people.
As for the circuit details, try page 18 lower left corner of this file.
The battery is down in that corner. This is an Intel reference schematic.
http://www.intel.com/design/chipsets/designex/BXDPDG10.PDF
In that schematic, two BAR43 diodes are ORing power together. The
higher voltage of the CR2032 battery voltage, or the "3VSB" rail
derived from +5VSB, is used to power the "RTC_BAT" node. If the
"3VSB" has a higher voltage than the battery, then the BAR43 next
to the battery will be reverse biases. If the +5VSB on the computer
is removed (by switching off the supply or by unplugging it), the
battery becomes the sole source of power. The BAR43 on the battery
is then forward biased and current can flow. Current can never
flow backwards into the battery, and the motherboard cannot
charge the battery because of that diode.
The 1K ohm resistor seems to be there, to prevent the battery from
being shorted to ground. A resistor like that, causing a high
source impedance, can be present, because at the time that the
+5VSB is removed, the computer isn't running, and there are no
accesses to the CMOS well on the Southbridge. Because there
are no accesses to the CMOS well, there is no "dynamic" current
flow. Only a very tiny current (a few microamps) is required to
run the ripple divider and digital counters on the RTC.
The voltage drop across the 1K resistor is minimal under
those conditions (a few microamps times 1K ohm).
When the computer is running, the BAR43 above the battery is
reverse biased and the battery is not an issue. Notice there
is no 1K resistor in the 3VSB path. That allows more significant
dynamic current flow, when Windows goes to read the RTC and
copy the current time into system memory.
Now, when you look at that circuit, you're going to say
"but Paul, how does the circuit get damaged ?". This example
from Intel is designed properly. It doesn't have a design issue.
The circuit would not be damaged, if you left the computer
powered and played with the jumper. But designers other than
the staff at Intel, seem to feel it is fair game to short
the RTC_BAT node to ground. With no current limit resistor.
In other words, the affected motherboards, the ones that can
be damaged, use a minor variation of that circuit. And those
are the ones that get damaged. The BAR43 in the upper leg, is
the component that burns.
On all the motherboards I own here, the two diodes are contained
in a three-legged transistor-like package, so in fact both diodes
are damaged at the same time, because they share a common plastic
package.
One person, who burned his dual diode, was able to
solder a couple 1N914/1N4148 type diodes in as a substitute.
I didn't recommend that, but that is what the guy had on
hand, and his computer worked properly after the repair.
Silicon diodes like 1N914/1N4148 have a higher forward
voltage drop than the Schottky diodes shown in the
Intel schematic. But as long as the circuit works
when the soldering is done, that probably doesn't
matter that much. The battery discharge curve
is sharp enough, that it probably doesn't affect
perceived battery life.
This is the dual diode package used on my Asus motherboards,
for that part of the circuit. Asus uses a BAT40W-05, common
cathode diode package. "K45" is the marking on top of the
device (i.e. marking is not the same thing as a part number),
and you can typically find that tiny thing, near
the CMOS battery. When it burns, the legend on the
top is no longer readable.
http://web.archive.org/web/20040229224112/http://www.diodes.com/datasheets/ds30114.pdf
HTH,
Paul