Rewinding an encapsulated SMPSU transformer.

[McMurdo]

I've just destructively de-encapsulated a burned-out mini SMPSU transformer. (oxo-cube sized). It was in hard epoxy, so chipping it open broke the ferrite E- core into tiny bits but left the bobbin, allowing me to back-wind it.

My question is, can I identify the core grade? I've found the 'series' of core/bobbin sets in Farnell. The bobbin is fine but the ferroxcube E-cores come in many different grades. I guessed at one and wound a new transformer. It worked for about 5 seconds then the SMPSU circuit popped and let out the smoke.

These little transformers fail usually due to a shorted chopper transistor. I have a good sample (encapsulated!) and have made inductance measurements.

[MrBungle]

There are also gapped and ungapped ones. If it's any consolation my attempts to build a switching converter with E cores all ended up in small fires. Either I blew the MOSFETs out, smoked the BJTs or toasted the coils. All while the coils screamed of the ferocious agony of their demise (they make quite a bit of noise if you fudge your calculations as they act as voice coils). My wife persuaded me to give up in the end.

[Oldcodger]

some time ago, I asked about ideas on winding a ferrite transformer for a home brew 12-240 SPMU. Suggestion was to use a transformer from an old laptop PSU.Might be an idea to look at.

[kalee20]

Like Mr Bungle says, there are gapped and untapped cores. It depends how much inductance, and at what maximum current, the circuit requires of its transformer.

The ferrite itself is comparatively much of a muchness. Different grades will give different temperature rises, but there generally isn't so much difference between them that a circuit will work with one grade and not with another.

[cmjones01]

What's the topology of the power supply? If it's an Oxo-cube sized transformer, it's almost certainly a flyback converter, so the core would normally have a gap in it to prevent saturation. Using a non-gapped core could well result in the core saturating, which would lead to uncontrolled current rise on each cycle and...a puff of smoke after a few seconds.

Try again with a gapped core. Even using a non-gapped core and grinding half a millimetre or so off the middle leg can make a difference. If you have the chance to run the primary side of the PSU from a reduced voltage and watch what's going on with a scope, you can make an assessment of whether it's happy or not before everything goes up in smoke.

Other things to watch for: shorted turns in the transformer, which very often happen after overheating due to saturation, and shorts in the secondary rectifiers or capacitors.

At this point we're in redesigning-the-power-supply territory rather than just a repair, but it's still an interesting exercise!

Chris

[McMurdo]

Yes. This is all good advice taken. I've read-up on the design and it's an EFD20 core with a self-oscillating RCC circuit which requires a gapped core . In fact the circuit appears to be the actual one in the textbook...90-250V AC supply at around 14V, with a 5V linear regulator hooked on the output. Even the chopper transistor is the one in the book. (MJE13003)

I've ordered a pair of cores and a couple of bobbins and will try again. If it works, I can make some more, as I have quite a few faulty ones.

I've run a good one on the scope so I know what it should look like.

A good example takes so little current that it can be run with a series lamp limiter of only 15W. Unfortunately that was still too much to avert disaster on my first attempt.

[MrBungle]

I wouldn't use a lamp limiter with a switching supply. It is a non linear device and you can end up with an accidental and quite destructive oscillator on your hands. If you have a lab supply that kicks out enough volts or two or three that can kick out 100v in series it is probably worth firing it up on that first with current limiting. The front end of the supplies is usually a rectifier and they will run off DC fine usually down to about 90v

[McMurdo]

I dont have a DC supply that'll go up that high.
I've always used lamp limiters when fixing SMPSU's, over 20 years experience. Never had one take a dislike to it yet, including 80's TVs with a 100W on the mains and a 40W on the HT line.
The only tricky ones are those with an auto switching voltage doubler (getting rare now) which can't make their mind up if they're on 240 or 110.

This particular smpsu is perfectly happy on a lamp as I've fixed loads that way. When the transformer fails it's usually blown the 200mA inline fuse and a heap of surface mount. Changing the SMT and dried electrolytics often gets it going but sometimes the lamp glows then you know the transformer is u/s too.

[McMurdo]

Update to this, managed to get it working.

I wound a primary first, then tried it for inductance with the core. Result; inductance way too low; about 7mH when it should be 34mH. After much head scratching, by experiment, I altered the number of turns until I got the right inductance; from 336 turns up to 440. Actually I put 600 on then unwound it; as it was easier! I wonder if I'd somehow miscounted the original when I unwound it, as it was badly burned and I had to keep stopping the winder when the turns snagged.

The feedback winding went on next, that measured spot on at 23 turns. Then finally the secondary at 20 turns, which read low inductance, so I added 2 extra turns to bring it up. I tried it, and it worked first time! Secondary volts across the capacitor were a bit high, 7.8V as opposed to a target of 6.8. Removing the extra turns brought it back down.
It's been on about 15 minutes so far, everything cool and taking 1.3W off the mains at 240V; same as an original one.

[kalee20]

Good result! Especially as, a phasing error and your switching transistor is toast.

Presumably the ancillary items, like inter winding insulation, screens, etc have all been incorporated in the rewind.

If it's survived 15 minutes, it'll probably survive thousands of hours. If there was anything wrong, you'd have known it by now!

[paulsherwin]

Just a quick warning to anybody coming to this thread via a Google search or whatever: it is extremely dangerous to attempt to repair a switch mode power supply by trial and error like this unless you know exactly what you're doing. These are safety critical subassemblies and it's easy to end up with something that is electrically dangerous or a fire risk. If you don't have the right training and/or experience, don't attempt a component level repair - just fit a new PSU.

[McMurdo]

Indeed. I wouldn't attempt this on any consumer equipment particularly if it's a class-2 device. Items with a small psu like that would seldom be worth repair at all! In this case, there's no interwinding screen as it's only for an on-board internal supply within an industrial instrument with no external connections or contact with anything that requires isolation. For the record, most wall-wart or phone supplies dont have interwinding screens either; just a few turns of polyester tape between primary and secondary.

Visitors to the youtube channel 'Bigclive' can see the effects of faulty interwinding insulation on a direct-from-china USB power supply.