Normally, it’s easier to just replace a PC power supply than it is to fix one. However, some PC manufacturers produce custom cases with custom power-supply enclosures. This scenario forces you to have to order a costly replacement directly from the manufacturer. In many cases, an hour of time invested and $10 worth of parts can save you $30-$100 that you would otherwise spend on a new power supply. Here’s how you can test the components inside of your computer’s power supply.
Danger!
House current is very dangerous and claims a number of lives every year. I will include safety instructions along with each debug step I offer. If you stick with the method discussed, you won't get hurt. If you plan to deviate from my suggestions, you're at risk of serious injury.

What can go wrong?
In my experiences, nearly all power-supply problems occur in either the high-voltage section of the power supply or in the voltage regulator stage. This makes a great deal of sense since these are the circuits closest to the external world. The high-voltage section transforms the house current from the wall socket to the lower voltages needed by the rest of the power supply. The voltage regulators are the last stage of the supply and connect the power-supply outputs to the various other components within the computer.

Testing the high voltage of the power supply
The power supply’s high-voltage section transforms house current into a lower-voltage alternating current that can easily be filtered and distributed to the PC. You can see an example of a typical schematic in Figure A. This schematic will not exactly match your power supply, but it illustrates the typical configuration. The graphic shows the typical arrangement of the various components and lets you see how each component might fail and ultimately cause power-supply failure.

Figure A
Here are the basics of what a power supply looks like.


Looking at the schematic in the figure, you can see that if any of the capacitors were replaced with dead shorts, this would cause the fuse to blow. Likewise, if the metal oxide varistors (MOVs) fail in the same manner, they would also cause the fuse to blow. Now, if these components were to fail and become open circuits, nobody would ever notice. These wouldn’t cause a catastrophic failure. Likewise, if an inductor became a dead short, again you might not realize it had failed. However, if the inductor burned up and became an open circuit, this would render the power supply inoperative. You can check for each of these conditions using your multimeter.

To use the multimeter for component validation, you need to set it to measure resistance. Most of the cheaper multimeters have a number of different scales for measuring various ranges of resistance. The more expensive meters will automatically scale. If you have one of the economy multimeters (as I have), set it to measure on the lowest scale. On my meter, this is the 100 ohm scale. The meter is most sensitive when set on this scale. If you measure across the fuse, you should see very close to 0 ohms measured. If you discover the fuse is blown, you must keep searching. A fuse rarely blows without something else having caused a more serious problem.

Likewise, you should be able to measure across the power switch and see infinite resistance in one setting and 0 ohms when the switch is pressed. If the switch doesn’t read 0 ohms in at least one configuration, you know the switch is defective.

When you measure across a capacitor, you expect the resistance to be very high. The only exception to this rule would be for C1, which might read falsely low because the multimeter will also see the parallel circuit looking into the transformer primary. If C1 reads 0 ohms, remove it from the circuit using a soldering iron and test it again. If it reads something on the order of 10 to100 ohms, the meter is probably picking up the parallel circuit formed from L1, L2, and the transformer primary. If you exhaust all other possibilities, pull the capacitor and test it.

Testing the inductors is much easier. Again you measure the resistance across these components and hope they are very low. If they measure anything short of infinity, then everything is good.

The metal oxide varistors should work exactly the opposite. If you read anything less than 100 ohms, the device is damaged. Normally, if these devices are damaged, the fuse will also be blown out. These devices will fail if a prolonged surge strikes the power supply. If these are found to be defective, you can replace them, replace the fuse, and then put the power supply back together. Almost always, these devices fail early enough to protect everything else. Technically, you could just remove the MOVs and test without them. If this is the only problem, then the power supply will function, but it will have no surge protection.

The last component in the high-voltage section is the transformer primary winding. This is just a coil of wire, as far as the high-voltage circuit is concerned. You can expect it to have a fairly low resistance when measured with the multimeter. If the resistance measures very high, say more than 100 ohms, the transformer has burned out and must be replaced.

At the bottom of the figure, you can see a full wave rectifier circuit constructed from four diodes. This circuit might actually be four separate diodes on some power supplies, while it might be a single device with four leads on others. Regardless of how it appears, it’s very easy to test. If you connect your multimeter across any diode, you should see a very low reading in one direction and a very high reading when you reverse the probes. If you get a very low reading regardless of which way you connect the probes, then the diode is bad.

In the case of the single package rectifier, the same rules apply. You can imagine the diodes being between neighboring leads. If you get low resistance in both directions, the device is defective. When the rectifier fails in this manner, it will sometimes burn up the transformer and will always blow the fuse eventually. If this happened, it’s very fortunate since the failure would probably not propagate beyond the rectifier.  If you had a burned-up transformer and the rectifier was still intact, the problem might have been within the internals of the power supply.

After making all the appropriate repairs to the circuit, check carefully to make sure you didn’t accidentally drop any solder blobs on the circuit board. If you replaced any components with new ones, make sure you trim the leads short so they won’t make contact with anything else, especially the metal cover.  Once you have checked everything carefully, reinstall the power-supply printed circuit board within the power-supply enclosure. Don’t install the power supply back in the computer just yet. You will want to test it out before going through the trouble of putting it all back in the computer.

When the enclosure is properly restored around the power supply, reconnect the power cord and inspect to see if the fan has spun up. If the fan functions, check the output voltages using the multimeter set to measure voltage. Look for a 25-volt or 50-volt scale on the meter. Measure the voltage between the black wire and the red wire on any of the peripheral connectors (four wires: two black, one red, one yellow). If you see 5 volts on this connector, measure the yellow wire to see if it carries 12 V. If both are present, you can install the power supply back into the computer (after first disconnecting AC power from the supply).

Testing the voltage regulators
After fixing the high-voltage section, if the +5 V or +12 V outputs are still not working, you have one last easy debug procedure to perform. You will again need to remove power, remove the power-supply enclosure, and locate the voltage regulators. These parts always have three leads, and most every power supply I have seen uses the very same part labeled LM7905 for the 5 V regulator, and LM7912 for the 12 V regulator.

It is possible that the power-supply manufacturer chose some other part for voltage regulation. If you cannot find the LM7905, try following the red wires back from any of the peripheral connectors. The land pattern on the printed circuit board should flow directly from this wire to the voltage regulator.

Once you locate these parts, you can choose any of several ways to determine whether they are working or not. Perhaps the easiest way is to just remove them and replace them with new ones you obtain from an electronics store. They cost under $1 each, so this might be the most efficient use of time.

Alternatively, you can remove the regulators and solder three wires in their place. Label each of the wires and make sure they are long enough to be accessible once the enclosure is again restored. This will let you inspect the input voltage going to the regulators when power is applied. To do this, you must again have the multimeter set to measure voltage on the 25-volt or 50-volt scale. With power restored, you then must measure the voltage between the ground terminal wire and the input terminal wire. The voltage measured across these two wires should be 5 volts or greater for the LM7905, and 12 volts or greater for the LM7912. If these voltages do appear, replace the defective regulator and the power supply will again function.

However, if they read zero, unfortunately the problem is outside the scope of this article. Power supplies are quite different in their inner workings, so it would be nearly impossible for me to write an adequate tutorial that would apply to most of them.

A place to start
Now that you know the procedure for identifying and correcting the most common PC power-supply failures, you should feel comfortable opening up and diagnosing some of the more common power-supply problems. The procedures I've outlined won't cure highly complicated problems; those are best left to technicians with the proper training.