Repurposed 3.2 KVA site transformer drawing 56A

[nick_theboatman]

I never did like transformer theory in my degree course and that was 40 years ago. I'm a boat builder now and we tend to have big battery banks and thus big chargers with all sorts of dodgy stuff going on. I have repurposed a series of power transformers to be used as the basis of high power (5KW) battery chargers. We have used things like old welder transformers and, more recently, 3.2KVA 240:110 site transformers.

We break open the yellow box, take out the trafo, strip down the laminations ("E" and "I"), rewind the bobbin (usually leaving the primary alone but invariably 160 turns (1.6mm diam wire) : 13 turns plus 13 (4 wires in parallel, 2mm diam)). The original windings are 160 : 40 + 40 and the CT is earthed. These seem to work well. We are finding that the original wire is enamelled aluminium but we wind with new enamelled copper.

The latest one is being problematic. The core was standard E & I but with all the Es aligned and all the Is aligned (ie not interleaved) and the joints between the E stack and the I stack were welded with a 1mm seam across the stack. I spent ages unpicking the core laminations then hammering down the little blobs (the residue of the weld). Re-wound, re-laminated (interleaved) and powered up for initial test. It seemed to work fine and produced the expected 19-0-19 output. But it started warming up. I then terminated the windings and powered up and it blew the fuse. I put on an ammeter and the ammeter said that it was drawing 56A off-load. I suspected that I had eddy current losses going on so I stripped the core down again and varnished all of the laminations (the corners where the weld had been were now shiny steel). Re-assembled, powered up to test and it blew the fuse again. Stripped the core down again and re-assembled not-interleaved and re-welded the joints between the Es and the Is. Powered up and again drew excess current, this time 62A. And blew the fuse.

Suspecting a shorted turn I cut the weld, stripped the core down again and unwound the windings (looking for burns or other evidence of shorts).

Not a sausage.

I am a bit perplexed that a transformer core, laminated so as to prevent eddy currents, is welded so as to short all the laminations thus promoting eddy currents...

But most of all I cant get my head around what is causing this massive excess current.

Any ideas?

 

[Lucien Nunes]

Was the original number of turns per volt fairly standard? Is it possible that it has been wound with insufficient turns or not all the lams and is saturating the iron? Was this originally a normal site transformer or something special?

One useful telltale is that if the main problem is eddy currents, the iron will probably heat faster than the copper, but if shorted turns the iron won't heat excessively but the copper will. Welds across lams work because their cross-section is small relative to the surface area of each lam.

 

[Simon47]

The welded core is a "cost reduction" measure.
You'll be well aware the effort in properly assembling staggered E&I cores. Imagine how much quicker it is to stack up a few E cores, slot on the bobbin, drop the stack of I cores in the jig - and let the robotic welder zip up both sides.
When your primary competition tool is price (and something like efficiency is a long way down the lisy), whacking them out of a low skill factory in Asia is the way to go.

 

[nick_theboatman]

The welded core is a "cost reduction" measure.
You'll be well aware the effort in properly assembling staggered E&I cores. Imagine how much quicker it is to stack up a few E cores, slot on the bobbin, drop the stack of I cores in the jig - and let the robotic welder zip up both sides.
When your primary competition tool is price (and something like efficiency is a long way down the lisy), whacking them out of a low skill factory in Asia is the way to go.

Agree with all of that. I guess that that part of my question related to "so if you deliberately short out the laminations (that'll be the laminations that you took so much care to isolate...) then how much additional loss might you expect, in eddy current, when welding the joints. I must say that I only have myself to blame: 56A coming from eddy currrent losses..?
Since I posted this I have stripped the thing down again and unwound the windings (meh) and no sign of shorts. However the winding wire is enamel over aluminium (and wasn't varnished so could be easily unwound) as opposed to enamel over copper and this stuff isn't great at being recycled. Also my winding wasn't as tidy as it might have been and so when I seated the over-large windings down into the bobbin with a mallet and wooden block I might have chafed some of the enamel. No sign of that either. Just ordering new wire anyway.

Was the original number of turns per volt fairly standard? Is it possible that it has been wound with insufficient turns or not all the lams and is saturating the iron? Was this originally a normal site transformer or something special?

One useful telltale is that if the main problem is eddy currents, the iron will probably heat faster than the copper, but if shorted turns the iron won't heat excessively but the copper will. Welds across lams work because their cross-section is small relative to the surface area of each lam.

160T pri = 1.5v / turn and this sounds pretty recurring. I used Dick Kleijer's design tool here Coil and transformer calculator - http://dicks-website.eu/coilcalculator/ and that seemed pretty competent but always spoke of a primary in the order of 270 turns (less than 1v / turn). Interesting. By the way anyone know how to get in touch with him? He seems to know his transformer stuff. The original trafo was a bog-standard site transformer, nothing special.
Thanks for your comments about eddy currents: again obvious (but not to me :/ ) - it was the coils warming up and not the core.
So I'm going to re-wind using Dick's web-tool and more turns on the primary and lay the layers down much better (!) and see if that fries the incomer.
And then report back here
Thanks everyone