CMOS battery drains quickly

[magistertc]

This problem is based on my experience with an ASUS Z390 Prime motherboard, but it appears to affect other motherboards
as well Install a new CMOS battery, configure computer, turn off computer and disconnect from AC power or switch off SMPS.
Return somewhat later and find that battery is drained, dead, or weak. Clock is wrong, other settings lost, etc.

Examine the power 'sharing' circuit -- when the ATX 3V supply is on, the computer will function normally. The computer is also
configured to support the RTC and some volatile RAM (when SMPS is on). What you may not realize is that the 3V coin cell
uses this same circuit to provide voltage to RTC when ATX is off.

This is done with BAT54 dual diodes, presumably Schottky type, two inputs joined to a common cathode output.
The RTC gets operating current,in a small amount, and the ATX components are left alone.

I have found that BAT54 can fail by allowing minor leakage from the CMOS anode through to the ATX anode. Result is
CMOS battery sending current to ATX components. You can check this by removing the coin cell, leaving SMPS on,
measure voltage at the positive terminal (or at BAT54) coming from the ATX 3V line.

Further proof -- remove the installed coin cell, install it in a floating coin cell holder, wire its positive terminal to a schottky
diode anode (such as MDR 160 100V 1A,, I happened to have a few in my parts box). Attach cathode and negative
leads to those 2 x 1 female front panel connectors, Plug this into the P1/P2 RTC reset terminals on the
motherboard. P1 is the RTC connection, P2 is common ground. This will bypass the BAT54 component.

This should maintain battery voltage as expected. NOTE: motherboard schematics on the web may be worth examining.

[stj]

is the reverse protection diode to the 3.3v line failing or the one to the battery failing?
if it's the one to the battery then it may try to recharge the battery!!

[piernov]

I like measuring current flowing out of the coin cell first in µA range, it should be a few µA, maybe 16µA max. If it measures higher you know there's indeed too much current being drawn out of the cell and it's generally one of 3 things: bad southbridge, bad SIO (it often has an input to measure the voltage) or leaky dual diode, any liquid damage or dirt (or previous battery leakage, it's very rare with CR2032 but it can still happen) around these can also be the cause. It could also be the capacitor but I haven't seen that yet. After you fix the problem you can measure the current again to confirm it's down to a sane level.
I've already seen leaks at like 100µA, which is not obvious but you'll still see the voltage going down more than usual after a few days of being unplugged.
The specific diode part number is BAT54C btw, other suffixes means other arrangement of diodes (although you might see that on machines with rechargeable coin cell batteries such as Toshiba laptops).