One example of a power supply repair

I should say right away: I fixed the original fault, but in the end nothing came of it.
Still, if you're interested in reading about the repair methods and the finer points, I think you'll find something useful and interesting here.

Here's the situation. There were two large air conditioners at a site. Both failed after a nearby lightning strike. Someone managed to repair one board; the other was brought to me after an unsuccessful repair attempt.
When a repair like this succeeds, I usually charge double for my work, because repairing something after someone else has worked on it is much harder. Today I'll show you why.
This is what the board looked like originally. Don't be surprised that there's no input filter capacitor on the board: it's connected by wires, and there's a connector for it on the board. This is fairly common in air conditioners.


But what I found underneath was even more upsetting. Any repair starts with a visual inspection, not with trying to power the device on. Never try to power up a power supply without inspecting it first—this is important!


There are clear signs of soldering on the board, and it looks pretty awful. That's exactly why I don't like repairing devices after someone else has worked on them: you don't know what was there originally or what was done. But let's see if we can figure it out.


1. The PWM controller and optocoupler have been replaced. By the way, this uses a PWM controller that doesn't need an auxiliary winding on the transformer. That's very convenient; I use these myself, but when I made a video about the subject, I forgot to mention them.
You can also see that someone tried to replace the Zener diode (on the left) and removed the second Zener diode (on the right).
2. The soldering is pretty dismal. I can understand a board looking like this after a component has been removed, but this is unacceptable after one has been installed.
3. You can also see that the bridge rectifier was resoldered, and a resistor next to it was removed. Its value is unknown.
4. There are soldering marks here too. But what surprised me was that someone removed the two-winding input choke. Why is a mystery. You can easily check the continuity of a choke like this while it's still on the board.


As I said, the most important thing is a preliminary inspection. It often lets you identify the cause of a failure and figure out how to fix it before you even pick up a soldering iron or tester. Let's say you can spot around 70-80% of problems by eye, without any instruments. At least with power supplies :)
There are signs of arcing near the transformer, along with metallization of the board caused by it.


I desoldered the transformer and found that an arc had formed between two pins. Since an arc sprays metal from the contacts as it burns, the board is covered with a thin layer of metal.


A visual inspection showed that the wires to the transformer's center pins had burned off. I free the remaining wire with a knife, then tin it. After that, I take a couple of short pieces of wire and solder them to the pins.


After failures like this, you often get a short between turns. I connect the transformer to an inductance meter and find that the primary winding inductance is about 1.3 mH. That seems about right. For comparison, in the second photo I've shorted the secondary winding, and you can see that the inductance has dropped significantly.


But don't rush to solder the transformer back in. As I said, when an arc burns, metal evaporates from the contacts—in this case, from the winding wire. This metal settles on the circuit board and will most likely conduct current. Best case, the power supply will simply blow up; worst case, it will become unsafe.
By the way, the power supply in some air conditioners isn't galvanically isolated from the mains, so in this case the only problem might be having to repair the power supply again.

In any case, clean the board thoroughly, and while you're at it, clean out the component mounting holes.


Always make the first power-up through an incandescent bulb. Don't use LED, CFL, or similar lamps.
The bulb's power rating is usually chosen based on the power supply's output. For low-power supplies (10-40 W), a 15-25 W bulb is enough; for 40-100 W power supplies, use a 40 W bulb, and so on.
On my first startup, the 15 W bulb started flashing in time with the relay on the board switching. After I replaced it with a 25 W bulb, everything was fine. Apparently, the board has a high power consumption of its own.
And don't connect a load during this test—the power supply is checked with no load.


During testing, I found that the 5 V regulator was getting very hot. In the end, I desoldered it from the board and, unfortunately, damaged it in the process. I then replaced it with a standard 7805.


I usually do this check before repairing the power supply, but this time I did it the wrong way around: I repaired the power supply first and only then started checking the rest. After desoldering the regulator IC, I applied 5 V to its output point. It turned out that the board was drawing 200 mA, which is why the regulator was overheating and shutting down the output.
The diagnosis: the microcontroller had failed. It was getting the hottest, and given that the board used a 78L05 regulator rated for a maximum current of 100 mA—and regulators are normally chosen with some headroom—while the photo shows the board drawing twice that much, the conclusion is clear.
Instead of the expected 50–70 mA, it was drawing 3–4 times as much.


Next, I decided to at least get a little closer to finishing the job, although there wasn't much point, since I didn't have a replacement microcontroller anyway.
But I wanted to show what to do if everything else is still intact, because the power supply itself had been repaired.

I mentioned above that one zener diode was missing from the board. It was part of the voltage regulation circuit. I knew the required voltage right away: that circuit powered relays marked 12 V.
I installed a 9.1 V zener diode, but it turned out to be too high, and the voltage was 16 V instead of 12 V. That's not a serious problem, but it was better to replace it with another one. I later installed a 6.2 V zener diode, and the voltage returned to normal.


Then I desoldered the sockets holding the PWM controller and optocoupler, since sockets aren't recommended in high-voltage circuits.
The procedure is simple: desolder the sockets (or old ICs), clear the holes, wash the board thoroughly, install the new components, then wash the board again.


I also tidied things up a bit on the underside. After repairs, especially on an air conditioner, humidifier (or dehumidifier), or washing machine, I usually coat the board with protective varnish, since these devices can be exposed to moisture. I use Plastik-70 varnish; its advantage is that it can be removed with acetone. If you want a permanent coating, use Urethane varnish.


That's about it. Today I covered a little theory and showed that you can repair a power supply, but still end up unable to repair the device itself. What a shame :(

And, of course, here's the video. This time it's about using a light bulb when repairing and diagnosing power supply faults.
In brief:
1. If the bulb stays lit continuously, there's probably a short in the input circuitry, such as the bridge rectifier, input capacitors, or power transistor.
2. If it glows at half brightness, one of the bridge rectifier diodes is probably shorted.
3. If it flashes at a rate of 0.5–2 Hz, there are probably problems in the secondary circuit and the PWM controller is restarting. The same thing can sometimes happen if there's a problem in the PWM controller's power supply circuit.