Sorry to hear that mate...
Well what I would do is to try powering different power lines with a lab PSU set up for correct voltages with the current limited to, say 1A, if you have one... That is the way I troubleshoot short circuits inside chips. But you need a lab power supply which is not a cheap toy.
If you manage to get hold of one, you might try powering the 5V circuit of the P8-P9 power connector with 1 Amp current and see where the boards heats up - that would be the short circuit place. Most probably inside one of the chips, but might well be a capacitor or a diode, or a transistor that gates something.

Having said that, I am not sure that there are no other ways to test that, this is just what comes first to my mind.

BTW, did you check other power lines, +12/-12, -5? 500 Ohms on 5V gives the current of 0.01 Amps, I do not see anything wrong with the number to be honest.

You certainly got something faulty but sadly it will take some real time and you will need equipment to hunt down the fault. I would check all the small logic first.

I would try to use a diagnostics card to narrow the source of error.

> I would try to use a diagnostics card to narrow the source of error.
Very useful with a shorted board...

> I measured the resistance from 5V to GND and it was around 500 ohms.
Also measure the other voltages.

Hint: They dont look like the typical ones, but the board has tantalum caps.

I will measure the -5V, -12V, and +12V and report back. I was going to do this the other day but became burnt out. Is it possible, though, that the should could still be on the +5V line but manifests itself only in the ON-state?

Do tantalum caps short before they explode?

I have a PCI/ISA diagnostic card but did not try it. Wouldn't it need an external source of power to display the error code?

GPA: I have a variable DC power supply with a variable amperage knob. I think the unit goes up to 30 V, 5 A, however it does not output negative biases. Wouldn't using this approach be the last resort? If the issue is a tantalum cap, couldn't I be provoking a fire?

There are quite a few tantalum caps on this board. Should I just replace them all as a starting point? I know there have been a few topics on this, but do you remember the rule of thumb for replacing tantalum caps? Replace with ceramic, mylar, or film caps of the same capacitance for a board of this vintage?

feipoa wrote:

I have a PCI/ISA diagnostic card but did not try it. Wouldn't it need an external source of power to display the error code?

I don't know. My card gets its power from the slot. It also has diodes for each of the voltages that flow through the slot connector.

Jo22 wrote:

feipoa wrote:

I have a PCI/ISA diagnostic card but did not try it. Wouldn't it need an external source of power to display the error code?

I don't know. My card gets its power from the slot. It also has diodes for each of the voltages that flow through the slot connector.

I do not believe the slots are being powered because the PSU is entering a shut-off state.

Try pulling the cache chips. They may be from a questionable source in China, and it wouldn't surprise me if one developed a short. At least, I saw this kind of thing happen with the Chinese bought 128kx8 SRAM chips.

Anonymous Coward wrote:

Try pulling the cache chips. They may be from a questionable source in China, and it wouldn't surprise me if one developed a short. At least, I saw this kind of thing happen with the Chinese bought 128kx8 SRAM chips.

Tried that already.

feipoa wrote:

...I tested all the SRAM,...

maybe you need to check the power socket on the board, for any corrosion and the like. Especially "power good" pin (where orange cable goes).

Some test results:

1) PC analyser card is useless because it is not getting powered.
2) There is a short from +12V to GND. The problem lies somewhere in there.

A) I am going to remove all the socketed components and start probing the caps for shorts.w
B) It is also possible to wire the power good line to 5V, run the system, and wait for the bad component to smoke. I hope it does not come to this.

feipoa wrote:

Some test results: […]
Show full quote

Some test results:

1) PC analyser card is useless because it is not getting powered.
2) There is a short from +12V to GND. The problem lies somewhere in there.

A) I am going to remove all the socketed components and start probing the caps for shorts.w
B) It is also possible to wire the power good line to 5V, run the system, and wait for the bad component to smoke. I hope it does not come to this.

Forget the PC analyser card, it is the CPU that issues the error codes under BIOS program, so it is useless if CPU doesn't run.

And I don't think that Power Good would work - it is the PSU that sais to the board that power is OK, not vise versa. So the PSU does not "read" PG, it "writes" it. "Writing" it on your own would not help to start the PSU.

12V on the other hand is good news! No chips are powered by 12V I believe, so the board should be ok. Some passives are probably causing the pain. Good idea to check capacitors. I don't believe there are pull up resistors to 12V, also chokes are not installed between +12V and Ground, so I would bet it is a capacitor issue. Some slot pins can be bent and cause shorts, but I didn't see that on photo.

feipoa wrote:

GPA: I have a variable DC power supply with a variable amperage knob. I think the unit goes up to 30 V, 5 A, however it does not output negative biases. Wouldn't using this approach be the last resort? If the issue is a tantalum cap, couldn't I be provoking a fire?

This is of course the last resort, but helps me out when I am dealing with damaged chips. If you can avoid using it, you should avoid it.
If you decide to use it, make sure you can limit the current, not only the voltage on your PSU. 30V 5A would provide you way too much current at 5V, and if the chip is shortened, it will drain all current your PSU can provide. This can end up with some nice fireworks. 1-1.5A is good enough for testing the shortened chips. You can test -12V by reversing polarity, applying +12V to ground and Gnd to -12V, given that you do not supply any voltages to any other inputs, it will work fine. But I cannot imagine -12V causing problems anyway, it is a thin and not so used rail, so simple resistance check is enough.

This fix was ultimately quite straight-forward. Thanks everyone for their input. After testing each through hole resistor and capacitor, I located a short at C100. To ensure that the short at C100 was due to the capacitor and not something on the other end of the traces, I desoldered the capacitor and measured its resistance. It was 0.1 ohms, or short. Unfortunately, I did not have any capacitors in the same package as the removed one, so I replaced it with a 25 V, 10 uF tantalum capacitor from a 486 board in my box of dead motherboard. The motherboard works fine now.

I'm glad I didn't set current-limited DC biases to the AT connector pins. This capacitor would probably have exploded. Even worse, I could have been feeling around the components for a hot spot and have the thing blow up or catch fire on my hand.

I am amazed how easily a single component can fail on these old boards. They really need constant maintenance to remain functional.

Here was the bad capacitor.

This is what I replaced it with.

That is the second board brought back to life recently! Congrats!

PSU would probably blow up the capacitor indeed, but it would not damage anything on the board, although would probably leave a mark on the pcb. Capacitors can be checked with a multimeter, but chips are a lot easier to diagnose with the PSU.

I've been getting lucky recently. I'm sure my luck will run out. I have a box of 20 dead motherboards for a reason...

Nice work! Also, I love seeing these stories posted 😀