An Experiment in Transformer Rewinding

I've had this transformer sitting in my junk box for a few years now. It came from a cheap hi-fi system. of (possibly) 80's vintage.

I'd worked out it must be rated 90-100VA, but with windings for 11.7v, 15.4v and 40.5v, it was a bit useless.

So when I needed a transformer with a power rating of at least 50VA to power a low voltage soldering iron at 24v, 2A, seeing as transformers are quite expensive to buy I naturally thought about re-winding this lump.

I established from looking at wire specifications that the correct wire gauge to use is about 18 SWG. The next thing I discovered is that a normal sized reel of this wire costs only a little less than a complete transformer. Whilst a smaller (read as, "hobbyist sized") reel is cheaper, it is much more costly in terms of £'s per meter. At this point I nearly abandoned the project.

However, also sitting in my junk box were two de-gaussing coils from old monitors. One I had previously unwound and rolled the wire up onto a cardboard tube. The other was untouched. I wondered if I could use this wire to wind multiple coils, and connect them in parallel to get the necessary current rating.

This is not an exercise in good transformer design! This method is almost certainly less efficient than winding the transformer with the correct gauge of wire. It is presented here as a way of making something I needed, with the materials I had available.

And so the experiment begins...

You need:

A transformer to modify
Wire for re-winding the transformer
Lubricant to get the last of the laminations back in
Insulation, ideally yellow transformer tape.
Terminations for your new winding

A chisel you don't mind damaging
A mallet
A bit of thin, strong steel
A vice

I found it very helpful to read this article: http://ludens.cl/Electron/trafos/trafos.html and this one: http://ludens.cl/Electron/Magnet.html

This is tricky.

The core is made of steel laminations in E and I shapes. If you are lucky, you get a transformer where all the E's are stuck together, and all the I's are stuck together. It's far more common however to find that they alternate.

Getting the first E out is the hard part. All the laminations will be stuck together and tightly wedged into the bobbin. You need to crack the adhesive and drive out the laminations without damaging them - particularly without bending them.

I used a woodworking chisel to split each lamination from the stack. I drove out the first one part way using a small steel ruler as a drift, then gripped the edge in a vice. I then tapped the transformer upwards on alternate corners using a mallet, so it came out in small stages.

After the first lamination is out, the rest are much easier as there is a space for them to go into when you crack the bond between them.

Once all the laminations are out, put them somewhere safe.

In this particular case, unwind 3 secondaries.

This particular transformer uses a split bobbin construction, with flying leads, making it very easy to dismantle. You may find you have one where the secondary is laid on top of the primary, in which case you can just start unwinding.

With a split bobbin, you get the primary and secondary coils wound onto separate bobins, which are stuck together in a plastic holder. A bit of prising and a few taps with the mallet, and the secondary bobbin came out.

Remove any tape and other insulation as you go. Save any useful looking bits of insulation, you can re-use them.

Knowing the voltage from each secondary winding, and by counting the turns from each of them as I unwound it, I was able to obtain a turns per volt figure by dividing the number of turns by the measured voltage. I actually obtained 3 silghtly different figures, so I averaged them and obtained a value of 4.26. To obtain a 24 volt output I therefore need 4.26*24=102.24 turns. A little experimentation to obtain the exact number may be required.

Carefully wind the wire onto something, since you may need to re-use it, for either re-winding this transformer, or for something else.

I spent a lot of time investigating this.

Wire is specified as having "ampacity". This means the current it can safely carry without getting dangerously hot. This figure is generally given for conditions of normal wiring, and is far too high for transformer winding. The reason this figure is far too high is because in a transformer, many current carrying turns of wire are tightly packed side by side, all generating heat in a parallel manner. Ampacity ratings are given for the wire laid out in a cable run, where it is far easier for the heat to escape.

Rather than get involved in a lot of maths involving current densities, I took the rule of thumb figure from the wikipedia entry on magnet wire of 2.5A per square mm. I need the wire to carry 2.08A, so it's sectional area needs to be .832 square mm. This gives a diameter of 1.03mm - the nearest standard sizes are 19SWG (in practice, 18SWG) or 18AWG. As previously stated, the wire quite expensive, so this is where the idea comes in of using many coils of thin wire which I already had.

I weighed and measured the wire I have, and concluded it is nearest in size to 26 AWG. The diameter is 0.4 to 0.45 mm, giving a cross sectional area of around 0.126 square mm. Since the current capacity of wire is directly related to it's cross sectional area, it's simply a case of seeing how many strands will make up the area I need. In this case it's 7 strands, which I can wind on as a minimum of 7 coils connected in parallel.

For current capacity, the thicker the winding the better, so I'll be adding as many windings as possible. I estimated that I have enough wire for 9 or possibly 10 windings, so that's how many I'm going to try to fit.

The amount of space the winding takes up needs to be considered. The amount of space on the bobbin I have is 17 x 11mm. This allows room for 1665 turns of 26AWG wire, however, due to the space taken up by insulation and wasted space at least a couple of hundred turns are easily removed from this figure.

The number of turns required is estimated as around 103 (given that I don't know the exact turns ratio, and I chose to round up rather than down), and with the correct wire gauge that would be it. However, with 7 windings, the total number of turns is 721.

Another factor that needs to be considered is the resistance of the wire. I measured the average turn as 17cm. Multiplying this by 103 gives 1751cm, or 17.5 meters. At approximately 138 ohms per kilometer, the resistance of this length is 2.4 ohms. Since I'm using 7 windings in parallel, this is divided by 7, giving 0.34 ohms. At 2A of output current, the loss due to this resistance is 0.7 volts - about 3 turns worth of voltage, so I'll add this onto the winding. Obviously there is a tradeoff here between the increase in resistance due to the extra wire, and the increase in voltage due to the extra turns, however for this purpose it's not important.

I wasn't going to wind hundreds of turns without checking that the voltage would be correct!

I wound the first coil with 104 turns and held it in place with transformer tape, since this was the figure given by the lowest turns ratio from the original 3 windings.

I quickly discovered that doing this in the living room with distractions of the TV and my other half is a very bad idea. I'd keep losing count after a few turns.

The ideal solution to the problem would be to mount the bobbin on a spindle with a turns counter. Lacking this, I used a permanent OHP pen to make a dot on every tenth turn - much easier to count when I lost my place!

This particular transformer needs to deliver 24 volts at full load. Since the winding is only at 1/7th of it's final thickness, only 1/7th of the load is needed, so I tested it with a load resistor consisting of 5 x 470 ohm resistors in parallel - not quite full load but it will do - actually more appropriate to 8 windings in parallel.

You can see in the two photos, 23.7 volts, which was the no-load output, and 23 volts, which was the output with a test load. It could really do with being a volt higher, so I'll add on another 4 turns, making 108 turns per winding.

As this is a "many windings in parallel" design, it now only remains to wind on the rest of the coils. A pretty boring and tedious job by any standard! I actually managed to fit 8 windings onto the core.

I took a photo of the second winding so you can see where I've made a dot every tenth turn. I did this because of numerous distractions, which caused me to keep losing count! At least with the dots I have a record of where I recently got up to.

I tried to start each winding at the place where the previous one finished, in order to keep it flat, however this plan began to fail on only the third winding, and I just had to fill odd spaces when I was able to.

I tested every winding using a partly reassembled core, to ensure each one produced exactly the same voltage. This is really important, a mis-match would lead to losses and heating! Good job I did, nearly every winding needed adjustment.

I joined all the start of winding ends together, and end of winding ends together, and connected them to flying leads. I used the pieces of card from the original windings to safely separate the soldered joints from the coils, before wrapping the whole thing in transformer tape.

You can see how the card works. First a wide piece to protect the windings. Next a narrower piece. The ends of the windings are hooked over this, so that if the leads get pulled, they are pulling against the card, not the winding. Finally a wide piece again to insulate it on the outside.

Put the core back together the same way it was built originally, slotting E's in from alternate sides.

Put in the 3rd and 2nd to the last pieces the same way, then you can slot the last piece between them, rather than up against the bobbin. You may need to file the edges a bit so it will go in. Mine was such a tight fit I ended up driving one of the I pieces in the other way to open up the gap, pulling it out when the final piece was part-way in. I squirted in some switch cleaner as a lubricant to help things along.

Slot the I pieces in, then tap it all together with a hammer. You don't want to see any gaps between the edges of the E's and the I's.

And there you have it. You can see the transformer connected to a 100 ohm load. With the full 2A load connected the voltage dropped to about 23.5 volts, which although not ideal, is adequate for my needs. Another couple of turns per winding would have been a good idea. The load (a soldering iron) gets nice and hot, and whilst the transformer laminations get warm - I suspect due to the iron loss having gone up due being dismantled and reassembled, but the winding stays nice and cool - just what is needed!

You can also see that the bobbin is quite full. I was wildly optimistic about how many turns would fit! If it was a single winding, the amount I reckoned on may have been more realistic.