the PSU does start by itself without being plugged into anything?
I don't know how far old AT PSUs are in regard to protections but it does sound like a short to me.

Post a picture of your board. If it has the "old style" tantalum caps, one (or more) could be shorted. If you have a multimeter check resistance of all rails vs. ground.

Here's the picture of the motherboard:

There was a Dallas battery but fortunately for me it was socketed so removing it was a breeze. These are the measured resistances in Ohms:

+5V (Power On) - cannot be measured, no value
+5V - 1510
-5V - 14
+12V - cannot be measured, no value
-12V - 14

They are weird values, so a short is probable. No tantalum or electrolytic cap seems to be broken (at least visually). Could start removing and checking bad caps on -5V, +12V and -12V rails.

usually the psu doesn't start when the power plugs are plugged in the wrong way, maybye it was plugged in the wrong way previously and the power connectors, power. rails are shorted or something is fried. Guess you need to give the power plugs and rails a good visual inspection, next continuity test of the rails then check the voltages under load.

something is shorted.... you can probably turn it on for 20 seconds cut the power and start touching the board and see what is "hot" which will be a shorted component but stressing it like that would be last resort

Ok, thanks for the advice. I'll see what I can do with it. Would be nice if I could breath a new life into it.

You can also try to plug in P8 and P9 separately and power-on. I wont hurt the board and may give you an idea where the error is located. Some old boards dont need P9 at all to work, dont know if its the case for a 486...

if there is no +12V and no +5V this is obviously a power delivery issue, nothing will start. You need to go and trace back where +12 and +5 have a possible common fail point and investigate what could of failed there

quicknick wrote:

Post a picture of your board. If it has the "old style" tantalum caps, one (or more) could be shorted. If you have a multimeter check resistance of all rails vs. ground.

+1 for shorted tantalum caps.
They either blow up when shorted or show no visible damage at all.
Check their resistance (not in continuity mode). If it's close to 0 Ω it's probably shorted. It worked for me in the past a few times, though not always.
Check the +12V and the +5V rail for short to the ground.

I would certainly be great if just caps were shorted. It's easy to replace them in the end. Hopefully it's not a problem with some if the ICs because then the board would probably be doomed. I will let you know if replacing caps helped or not, at least to have some kind of conclusion to this thread for future readers with similar problem.

I would like to write about a progress I have made with the board. While measuring it again I had found out short on -12V rail so I replaced a culrit cap and the short went away. The PSU now starts!

Unfortunately the board doesn't POST. I tried to hook up PC speaker to it but it doesn't even beep. My current rail resistances:

+5V - 1.505 KOhms
+12V - 1.5 MOhms
-5V - aaprox. 6 MOhms (keeps decreasing, probably ok as it is negative voltage rail)
-12V - 0.5 MOhms
+5V power - 6.36 MOhms

All has been measured with the Dallas battery out. Are any of the values unusual?

The most obvious thing to check first is jumper settings. Check the CPU orientation (if it was inserted incorrectly, then it's probably fried now). Some boards with a Dallas RTC will simply not POST with a fully charged battery. Try and find a replacement and see if it will then POST. Make sure you are using FPM RAM and not EDO RAM. If still nothing, then you possibly have a faulty BIOS ROM chip.

I see a lot of good suggestions here and I just want to add another.
.
If you found one bad cap there is a good chance there are others and not all caps are directly on the main PSU rails.
You won't see a shorted one of those by only checking the main rails.
To find those you have to check each cap at the cap.
.

I have measured all other caps and there is certainly no additional short, not on tantalum or elecrolytics. Some caps have infinite resistance between theirs legs - is that normal?

My jumper settings are fine, I have double checked them. I don't have any cache installed, could that be a problem? There are jumper settings for different cache setups, but no settings for no cache. So currently I don't have any cache jumpers set.

Could try reflash the BIOS if it's not corrupted. In my experience when PC doesn't POST, there is either hardware problem (damage) or badly set CPU or bad BIOS. With no beeps it's hard to narrow the problem.

I've put a new battery (probably fake one from eBay) in but that made no difference.

Paar wrote:

I have measured all other caps and there is certainly no additional short, not on tantalum or elecrolytics. Some caps have infinite resistance between theirs legs - is that normal?

You can only find shorted caps with in-circuit resistance checks.
(It's a sensible check because you are looking for shorts.)
It will miss other problems because other components will mask the other problems.
Even a cap that reads as shorted in circuit might have a shorted diode in parallel giving a false positive on the cap being shorted...
... however diodes failing to a short is very rare and shorted caps are common.

Out of circuit you can check both ways and watch the cap charge and discharge (on the meter reading) each time you switch polarity.
An anaolg meter is easier to see that with because with DMMs all you see is numbers jumping around.
Even that won't show everything.
The capacitance, ESR or leakage can still be out of spec on a cap that 'has enough left' to charge and discharge.

To fully check caps you need an ohm meter and a capacitance meter and an ESR meter (and technically a leakage tester).
(I don't bother with leakage tests myself. If the other 3 checks are good that will probably be too.)
That's a lot of gear for someone that only does this occasionally which is why many people that suspect capacitor problems just recap the whole board.

Personally if I have to pull caps to check them I just put new ones in. It's the same amount of work.
Checking the ones pulled only serves to tell me if I've cleared a problem cap or not.

Capacitance goes down and ESR up as caps age so caps over 15 years old are NOT operating at their original parameters even if they are technically still in spec.
For example, capacitance can change by up to 20% over a caps lifetime and still be in spec.
The people that think keeping the old caps on their boards so they have the 'same as original' experience or whatever are misinformed.
They are not getting the 'same as original' experience with old caps on their mobos, video cards or sound cards.
Just NOT.
.

When you say the motherboard doesn't POST, do you mean the screen stays blank when the motherboard is powered up? Or do you mean you see an image on the screen, but the board fails to pass one of the self-tests and therefore does not attempt to boot?

Do the keyboard lights flash at power-on? Did you try different graphic cards, preferable something simple like PCI Trio or Virge as well as an ISA graphics card?

What still confuses me about capacitors is why does it seem like I have rarely, or even never had to replace the capacitors in my ancient stereo electronics, but for motherboards, they all fail out within around 15 yrs. I drive an old Benz that I've had for ages, its a 1979. It has the original radio in it with a cassette player. Compared to modern car radios, I could one would say it sounds pretty horrible, and due to how recorded sound quality playback has improved a lot since the car was born, I do not really recall if my 40 year old stereo has been sounding worse or not over the decades. Why don't the caps in my stereo need replacing? I've thought about taking it apart and recapping it, but was concerned that it would just sound the same as it already dose.

feipoa wrote:

What still confuses me about capacitors is why does it seem like I have rarely, or even never had to replace the capacitors in my ancient stereo electronics, but for motherboards, they all fail out within around 15 yrs. I drive an old Benz that I've had for ages, its a 1979. It has the original radio in it with a cassette player. Compared to modern car radios, I could one would say it sounds pretty horrible, and due to how recorded sound quality playback has improved a lot since the car was born, I do not really recall if my 40 year old stereo has been sounding worse or not over the decades. Why don't the caps in my stereo need replacing? I've thought about taking it apart and recapping it, but was concerned that it would just sound the same as it already dose.

Good question!
Your observations are accurate and make sense, but to fully explain would be a huge hijack of the OP's thread.
It's not one reason behind the phenomena(s), it's several things working in combination.
My "vogons time" lately is about 10 minutes a day and that question needs a long multi-part explanation.
I will try to find some time next week and address it properly in it's own thread.
.

PCBONEZ is right, but those reasons most germane to this thread are capacitors on computer motherboards (1) require performance at much higher frequencies than car stereos, hence tantalum capacitors in some cases with their service issues, and (2) subject capacitors to higher frequency noise than car stereos resulting in higher operational stress and greater failure rates.