I needed a small transformer for my next project, with 4.5 Vrms at 300 mA. I have dozens of old “wall-warts” and one was just begging to be hacked: it was already cracked open; probably got dropped on the cement floor. It was 9 Vrms nominal output, twice what I wanted. I considered my options. I could just use it as is, with a series power resistor, wasting power and generating heat. Or find a more suitable wall-wart and cut it open. Or buy a new transformer online or locally.
I decided I’d try to take it apart, remove about half the secondary winding wire, and put it back together. The transformer core is made of metal plates that look like the letter "E" and "I," arranged like this:
I used a razor blade to score the enamel between the plates. The first E plate was hard to get out. I used a piece of thin metal, to push it down, inside the plastic bobbin. It was in there pretty tight, but once I moved it an 1/8 inch or so, I could grab it from the other side with a pair of pliers and pull it out. The rest of the plates were much easier to remove.
That left me with the plastic bobbin with the primary and secondary coils. I carefully removed the plastic tape around the secondary coil, then removed what I guessed was about a third of the enamel covered magnet wire. I put the plates back in, fired it up and checked the output. It was still higher than I wanted, so I repeated that process a few more times, until I got the output I wanted.
A couple lessons: the inductance of the primary winding is very low without the metal core. At one point I fired it up, very briefly, with no core plates and it got very warm, very quickly. It would have burnt out if I hadn’t killed the AC power right away. So don’t do that.
Then I installed just the E plates, all facing the same way, without the "I" plates. That way I could easily remove a few turns off the secondary coil, test and repeat. But the metal core was open on one end, so the waveshape was distorted, and the peak to peak voltage was about 5% less than with the complete core. But the primary coil didn’t overheat, and it made it a lot easier to remove a few turns of wire at a time.
The yellow trace is with the complete core assembled. The wave-shape matches the input wave-shape (not shown). The white trace is with the E plates only, all facing the same way. The magnetic circuit is not complete and causes the wave-shape to be distorted. The peak to peak voltage is reduced by about 5%. The rms voltage is reduced about 18% percent.
Once I got the output voltage I wanted, I trimmed the copper wire back, scrapped the enamel coating off the ends of the wires, soldered on some suitable plastic insulated wires. I wrapped the secondary coil and leads with several layers of electrical tape. Since the plates still had a fair amount of enamel on them, it would hard to get them all back into the plastic bobbin. So I cleaned them up with a Dremel tool, barrel sander, and they all slid in easily.
All together this took about 2 hours, so not really worth it economically, but satisfying none the less.