Charlie Hoffpauir wrote:
On Mon, 01 Jun 2020 09:13:03 -0500, Charlie Hoffpauir
wrote:

My computer won't restart after a local power failure last night. Our
neignborhood lost power for 7 hours, but the computer was on a UPS
when we lost power. I shut it down with the power switch , but after
power was restored, it wouldn't start. I thought perhaps the PS was
fried in spite of the UPS because when the power went out, it didn't
go out all at once, it "flickered, that is, on and off rapidly for
several seconds. Anyway, I installed a old PS that I had used in
another computer, but no effect. Now I'm suspecting the on/off switch
on the case. There are 4 wires from the switch to a pin set on the MB,
red, black, white, and green. Can someone tell me if it's possible for
me to short two (?) of these wires to bypass the on/off switch, and
thus see if indeed it's just a bad switch? And it that's the case, a
way to simply bypass the case switch and get back in operation?

OK, I found out that it's just a bad power switch by unplugging the
switch wiring from the MB and shorting two of the pins on the MB, and
the computer starts up (on the bench, disconnected from everything).
so now I'm going to try simply adding an external momentary switch,
hoping that will do the trick.

I've had a switch go bad like that. Possibly a dome switch
with thin plastic cover.

Whether a switch fails permanently closed or a switch fails
permanently open, it's not going to work. The switch has
to have both positions working, for the logic to be
sequenced correctly. The Power switch needs both positions
to work. The Reset switch on the other hand, if it's jammed
ON, then the machine stays reset forever and will not boot.
The screen won't light up in that case. An indication this
is happened, is the NIC PHY doesn't autonegotiate with the
router and the NIC LEDs may stay off too if Reset is jammed ON.

You can leave your temporary momentary switch in position if you want.
The machine I'm typing on, there's a switch hanging off the chassis
right now :-) It's been like that for years.

Only two wires are needed for a single switch. If the
assembly has four wires, the plastics could contain
a Power switch and a Reset switch. Where the X's are in
the diagram, suggests how four wires could be arranged.
You would think, logically, that the ground could be shared,
and it only needs three wires, but they arrange the wire
in twisted pairs, and four pins works better with two
twisted pairs, during fabrication.

PWR#
+5VSB --- 10K ----+----X |
+== Momentarily closed for Power
GND--------------------X |

RST#
+5VSB --- 10K ----+----X |
+== Momentarily closed for Reset
GND--------------------X |

The switch bodies float with respect to chassis. Generally,
the switch body is not purposely connected to chassis, and
in a lot of cases the switches are set in the plastic facade
anyway. They like to make up twisted pairs for each
switch (as a means to reduce capacitive transient coupling).

Years ago, we had a box at work, where the Reset line worked
as an antenna. If there was the tiniest ESD discharge into
the external metalwork on the computer, the box would RESET
on the spot. (I got a plane trip to observe customer
symptoms first hand. The machines were quite sensitive.)
My boss knew the layout of the wiring in the box (the managers
did the chassis, the grunts did individual printed circuit boards),
and he figured it out right away, that the two foot long wire
stretched across the chassis inside, was a bad idea. That's his
PhD in Physics cutting in.

I'm not aware of any similar situations on PC chassis, so
those twisted pairs must be working AOK. There are a few things
about PC chassis design that defy logic, but you can't argue
with success.

Paul