Winding an Isolation Transformer - Winding Design Question

[killingtime]

Hi,

I'm about to re-wind an 5 KVA E-I transformer for mains isolation. 240 in, 240 out.

It's currently wound as a 240-110 step down auto-transformer, so no isolation. I'll have to unbolt the laminations, remove the bobbin and unwind the lacquered copper. I can count the turns as they come off, and so long as I put the same number back on there should be no issue. Existing config draws very little magnetization current at idle.

1) My question is, do I go for a split bobbin (primary above the secondary - no overlap), or secondary wrapped around the primary? What's the trade off?  Later would be easier to wind.

I'll be putting a grounded copper isolation shield (taking care not to make a shorted turn) between the primary and secondary to reduce capacitive coupling.

2) Do I need to insulate the copper isolation shield from contact with any windings?  Thinking of using Kapton tape. The windings will already be insulated due to the lacquer.

It also looks like the existing config uses two 1.8mm wires in parallel for the auto winding, to increase the current handling capacity. If I'm careful, I should be able to split these and use one for the primary and the other for the secondary, assuming the Mfr hasn't lacquered the winding after the fact - in which case the copper will be junk. I read somewhere that re-using Xfmr wire was a bad idea. Any truth to this? Worst case is a shorted turn and I re-wind again.

Thanks. 

[Benta]

I don't think this will work as you expect.
An autotransfor is basically just a tapped inductor with a smaller core size than a real transformer.
But to answer your question: the best thing is to use a two,chamber bobbin for the windings to keep them isolated from each other.

[Ian.M]

As Benta has pointed out, autotransformers don't work the same as ordinary transformers.  Neglecting magnetizing current, the current in the bottom section of the winding is 180° out of phase with that in the top section and the supply, so they sum at the tap.  That means your 5 KVA 50% tap transformer, if you were to split its winding at the tap would effectively be a 2.5KVA 120V in, 120V out transformer.   Yes, you could rewind it for 240V, but at that point,  the only thing you are keeping is the core, as the copper is too thick to fit enough turns, and the bobbin needs to be replaced with a scratch-built custom one with a very well insulated partition. Also, hand wound with extra insulation and screening, you would be doing well to fit wire thick enough for 2KVA.

[killingtime]

...you could rewind it for 240V...

Yes, this is what I intend to do. I probably wasn't clear enough in the 1st post. I intend to completely remove the existing copper and use thinner copper but keep the same number of turns for a primary so the volts per turn is maintained. Then add another winding for the secondary. So there will be 2x as many turns in total (Pri+Sec) as is currently.

... but at that point,  the only thing you are keeping is the core...

Yes, understood.

...as the copper is too thick to fit enough turns...

Yes, understood, but only if I re-use the existing copper as is. If you re-read my 1st post I explain the Mfr used 2x copper wires for the whole auto winding to increase the diameter. They could have used a thicker gauge but didn't, they went with two smaller cables in parallel. As I'll be un-winding all this, I'll be able to split the double winding into two single windings - assuming the Mfr didn't soak the core in varnish. The same amount of copper is going back onto the core, but as two isolated windings and not one big one.

...and the bobbin needs to be replaced with a scratch-built custom one with a very well insulated partition...

Only if I go the split bobbin route, but I can 3D print that. I don't expect the winding temp to go above 100 DegC. ABS melts way above this.

...Also, hand wound with extra insulation and screening, you would be doing well to fit wire thick enough for 2KVA.

Yes. I knew the KVA rating would halve (at least) as I'd be winding a secondary on with the same number of turns, so the wire gauge would have to be at least half the size as the core window remains the same - reducing the KVA. I don't need any more than 2KVA. That's why I picked out a 5KVA to begin with. The sheer size of the core also means you need fewer turns compared to a smaller Xfmr to keep flux saturation at an acceptable level, so it's less effort to wind. A larger core can have a higher volts per turn, meaning fewer turns.

On the split bobbin arrangement .. what does this have over the overlapped arrangement as seen on toroidal Xfmrs? I see plenty of toroidal isolation Xfmrs where the secondary completely overlaps the primary. There's no split bobbin. Sometimes you also get a earth shield between the two for safety which I'll be adding anyway. It would be much easier to wind and I'd get to re-use the existing bobbin.

If I recall, split bobbins have poorer regulation, but better isolation. How much isolation is enough?

Regards.

[Ian.M]

Even though the existing copper is in the form of two parallel wires each of half the required CSA, its no good to you for rewinding because you need more room in the core windows to accommodate the inter-winding screen and extra insulation, so need wire of smaller diameter to get enough turns on.  Also the kinks in salvaged wire will make near impossible to rewind to the same packing density, and may compromise its insulation, increasing the risk of shorted turns or flashover.

The problem with layered windings on a single former is insulating between them.  The tape layer doesn't form a reliable bond to the sides of the former, so the effective creepage distance is only the distance between the windings round the edge of the tape layer.  To fix this you need insulating margins that reduce the width of each winding, costing core window area, so that the wire to wire creepage distance round the edge of the full-width tape layer is sufficient.  This does not apply to toroidial transformers because the tape layer overlaps itself at each end rather than butting up against a former side.

A FDM printed former is very unlikely to be suitable as its full of tiny voids which means it needs to be a lot thicker to get enough insulation as its liable to creepage through the voids resulting in tracking and total insulation breakdown.

[johansen]

I have put 8 kvdc across ethernet cable before (not the connectors)

A 1mm thick bobbin is sufficient unless you want to perform diy open heart surgery

[Ian.M]

That would be a 1mm thick solid bobbin, either moulded or built up from glued sheet material, with no voids in the glue joints.  Its extremely difficult to make FDM printed parts non-porous - see the leakage problems commonly found with 3D printed air, vacuum, or water hose fittings.

[moffy]

With regards to the original OP post, split bobbin should give better isolation, less primary to secondary capacitance and slightly more leakage inductance i.e. a bit more droop with load. Winding one on top of the other is of course the reverse. I would insulate the shield layer it would help with the isolation between primary and secondary if you go for one winding on top of the other.

[killingtime]

..its no good to you for rewinding because you need more room in the core windows to accommodate the inter-winding screen and extra insulation, so need wire of smaller diameter to get enough turns on.

There's already a sizable gap between the existing winding and the core. Looks like the Mfr didn't use all of the window. Might lose some turns, just means higher magnetic flux density and a larger reactive current at idle. Worst case I just start again.

Also the kinks in salvaged wire will make near impossible to rewind to the same packing density, and may compromise its insulation, increasing the risk of shorted turns or flashover.

I've got nothing to lose other than my time. The copper is already there. I've read you can pull enameled wire through a small hole drilled in wood to straighten it out again. Probably reduces the insulation though.

A FDM printed former is very unlikely to be suitable as its full of tiny voids which means it needs to be a lot thicker to get enough insulation as its liable to creepage through the voids resulting in tracking and total insulation breakdown.

I wondered about this as well. Others have been and tested FDM plastic insulation resistance. It's quite high. I remember reading something on PLA. At 100% fill it's very good. An FDM printed former wouldn't pass UL spec, but this is just for me and my bench. I'm not selling it. I could probably use the existing bobbin anyway, and just partition it in the middle with a PLA\ABS printed partition.

---------

Picture of the Xfmr is attached, together with a handy weight chart for different transformers.

I originally downloaded this from an audio website (can't remember which one), but I've added 3 different manufacturers to it since and shown the size & weight tables to save looking values from the chart. Useful if you don't have plate info for a Xfmr and want to know it's likely VA rating. Might be useful to others on here. Applies to two winding set-ups only, not auto transformers. Quite a difference between EI and toroidal KVA for the same weight.

My Xfmr weights 16KG with the copper - just weighed it. So that puts the cont. VA rating at 1.0 - 1.25 KVA.

What the hell... How did the Mfr get to 5KVA?

Low (1%) duty cycle, or running a high flux density?

I know auto Xfmrs use only 1 winding, so it can be twice as thick and therefore half the resistance -> about twice the KVA rating, but I don't see how they got to 5KVA. If anyone knows, please enlighten me.

Thanks.

[MarkT]

A 2:1 autotransformer carries 1/2 the power directly through and 1/2 through the magnetic field, so if the output is 5kVA it's effectively like a 2.5kVA transformer.

Using that core for an isolation transformer means its limited to 2.5kVA. Commercial 2kVA isolation transformers are about 10kg which agrees well with your 16kg at 2.5kVA, don't know where you got the 1 to 1.25 kVA figure from - transformer mass and power ratings are not linearly related as cooling is more efficient in a smaller device (or an oil-cooled device).

[Ian.M]

So, before destroying the original winding, measure the spare space in the core window as accurately as you can, and estimate if you've got enough to accommodate either two 2.5mm wide margins (for 5mm creepage distance), one either side of the winding the full depth of the former, + the thickness of the primary to shield insulation, shield itself, and shield to secondary insulation, or the extra material in a split former.

I wouldn't try to FDM 3D print a partition to convert the existing former to split.  Gluing two layers of thin plain FR4 together, each consisting of a U shaped piece and a rectangular piece between the legs of the U, inserted from opposite sides so the joints are staggered, with a fillet of glue either side where it meets the bottom of the former is much more likely to give a result you can trust.

Also, due to the thicker, narrower windings required by the need for margins or by a split former, the existing wire's length wont be enough to get the same number of turns . . . 
   

[johansen]

- transformer mass and power ratings are not linearly related as cooling is more efficient in a smaller device (or an oil-cooled device).

Transformer power follows the 4th power of the linear length of the cube at constant loss per kilogram. At constant watts per surface area they still have increasing power to weight ratio as they go up in size.

I have a 12kva transformer only weighs about 50kg, and is more efficient than OP's transformer at 2.5kva.

Anyhow, the number of people telling OP not to do this is crazy.

The transformer is not full of copper because copper costs 9 times as much as the iron does. OP can fill it up and report how hot it gets after 1 hour at 5kva.

[johansen]

My Xfmr weights 16KG with the copper - just weighed it. So that puts the cont. VA rating at 1.0 - 1.25 KVA.

What the hell... How did the Mfr get to 5KVA?

Its only 2.5kva as an auto transformer.

The charts you found are probably for 40C temperature rises. Basically what i said for constant losses per surface area.

You can find efficiency charts for toroidal transformers on toroid dot com. The 1kva cores only lose like 2 watts, the copper around 40 watts for a 96% efficient transformer that might weigh 8 pounds

Most of the stock toroidal transformers can be run at near double their "rated"  load for a significant length of time before they reach 80C temp rise.

Your isolation transformer could easily run at 80C temp rise and it just doesn't matter.

[killingtime]

A 2:1 autotransformer carries 1/2 the power directly through and 1/2 through the magnetic field, so if the output is 5kVA it's effectively like a 2.5kVA transformer.

Ah, OK. That makes a bit more sense.

... don't know where you got the 1 to 1.25 kVA figure from ...

The table for EI transformers from ETC I attached above (Xfmr VA vs Mass inc Copper.jpg). They make one that is 1KVA @ 16KG, and that's got two windings.

As johansen has said, I can run it at different loads and just see how hot the windings get. That's the ultimate test of the VA rating. Then just run it below that.

Thanks for all the replies.