Rewinding a shorted SMPS transformer

Not much problem with it, I do it simply as another part of prototyping.

Since you've already made transformers, I don't really have anything else to say. Insulation, wire size, all that stuff applies the same. Just avoid big fat wire, since it gathers eddy currents just the same as non-laminated iron does at LF. Multistrand and litz are wonderful things.

As far as taking things apart, use heat to soften up the varnish. Even so, cracks are common, which is why I say this: "Superglue is an integral part of disassembling ferrite cores."

Tim

Are the "coils" physically separate? Do you know if they were wound bifilar? Ferrites are more brittle than the iron cores you are used to with "60Hz power transformers"...

Just like one might do with power transformers: count the turns as you take them off, measure wire size and replace. Pitfalls: getting the damn thing apart without breaking - since it is that critical you will discover that no solvent in the world (save H2O which will dissolve everything) will get that damn lacquer / paint / epoxy off because it is all three and more insoluble than the core itself. Been there.

yeah...a LOT more brittle.

Hmm, I'm willing to bet 70% nitric acid would do a number on it. It might not eat away the core too quickly, but it will certainly be etched. Of course, you won't have any turns left to count, as the copper will disappear before anything else.

Tim

It is important to know why is failed. You are not ready to rewind=20 until you know why and how to prevent recurrence.

For someone with experience in winding mains transformers, the only major problem is taking the core apart without breaking the brittle material.

Some time ago, I had to build an SMPS from scratch for a client. The original was badly burnt and the company quoted >$2000 for a replacement. I designed a new one to meet the specs and at first thought I'd wind the transformer on the core of a transformer from an old computer SMPS. Taking it apart proved to be more of a hassle than I anticipated, so I just bought a new core and used that.

I didn't have to count the turns or measure the wire gauges as the new one was my own design. But if I had to, I could have done it by unthreading the windings through the core window or simply by breaking the core first.

(I felt adequately compensated by the $800 I charged my clients, and they appreciated the >60% saving in cost. A satisfactory deal all round).

The best way to determine winding construction on a vacuum impregnated pot core transformer is to source and buy a new core and bobbin, and then destructively dissect the old one. A cros-sectional cut through the bobbin of the old core will allow one to examine winding layers and get turns count and wire gauges from that.

Many transformer failures are due to poor or improper construction choices made either by its designer or by its builder. The failure mode is usually in the form of an arc-over event. If a transformer gets driven to failure through current, then the drive side of it was poorly designed, or the transformer itself was poorly designed electrically speaking.

This is, of course, a reference to small form factor pot core transformers.