shorted transformer winding

Reduced L, reduced Q, reduced R, heating if you drive it hard, maybe resonated.

If you have an IR viewer, you could drive the coil and see the hot spot.

Or drive it with a audio-frequency current and wave a small probe coil around. The pattern will be different on the bad board.

Measure the resistance of good vs bad boards too. Also check the resistance to see if they gave you real 2 Oz copper; lots of board houses skimp.

John

The blown board (has bulged spot where gasses boiled out to the surface) shows really strange L (I apply DC voltage and record current, L=3DV*dI/dt). The L value is piecewise, starts with 1/3 of what it should be and changes 0.8 of what L should be. Resistance is barely distinguishable (between burned PCB and "new" ones). 0.58Ohms vs 0.6Ohms. I have 20 turns winding. If I take out (short) one turn; it is 5% difference in R. If I consider that I have four layers, it means

The interesting part: I took another (brand new) board, inserted core, measured inductance. The waveform (I vs time) looks perfect, BUT inductance value (dI/dt slope) is 12(twelve!)% off. The core tolerance is 3%. Hmmmmm?? I swapped cores, I did make sure that cores' contact surfaces are clean and free of breadcrumbs, I said the prayer facing East, etc... Exactly the same result. What am I missing?

I presume that your board has got cut-outs so that the two halves of the transformer can be clamped together, otherwise you wouldn't be worrying about bread-crumbs, human hairs and the like.

-- Bill Sloman, Nijmegen

ed

It is gapped core. I assume that it all boils down to machining accuracy...

It is flyback....

Is a pcb-based transformer really worth the hassle? Buying a finished, tested part has a lot of appeal.

John

First off, 6 mils gap can be held in at-home etching with only moderate care being needed (ignoring possible undercut issues). Inductance and Q measurements would show up as lower inductance and lower Q.

Ramadan is not complete yet..

That makes more sense. The core gaps aren't "machined" but "ground". You do have to be very careful to keep the mating faces of the cores clean and well-clamped if you want to get the +/-3% accuracy you are paying for. You may need to worry about the stadning DC current in the coil - the permeability of the core changes with magnetic field, and almost entirely goes away if the field gets up to a few hundred Tesla.

-- Bill Sloman, Nijmegen

like

cing 2

al

ow

horted

ybe

coil

aybe

what

re,

es'

er

lean and well-clamped ... Done very carefully. I learned hard way years ago that epoxy may be used only in the gap, and surfaces that touch each other must be as clean as..."optical" surfaces (propanol cleaned and canned air dried). What would make inductance 10+% smaller than it should be? The test was repeated with two different cores.

Side story that expert may find entertaining. Few years ago I was looking around for an epoxy to put flyback transformer core together. My fellow worker gave me single use envelope of some stuff from his drawer. I mixed it, put transformer together and .... wondered why this stuff was silvery in color. I looked up part number. Sure enough it was silver filled conductive epoxy. It was not too late to clean it up... I thought of troubleshooting the circuit with flyback transformer having blob of metal in the gap... Ouch!

So, the rest (with the same pattern) are good?

You could also run the circuit, as intended, noting any local heating. Most likely location is at inter-layer routing, through-holes etc. These may even be visible through inspection, if within two layers of the surface.

If the circuit has a generous current limit, you might clear tha short, though I wouldn't use it for anything later, outside of the lab.

High current DCR can also be an indicator, even at low R walues; the length/width ratio of a short that's only 6mil long will still be suitably low, compared to a full turn. You'd get nearly the same type of effects as AC testing, though not as definitive.

RL

and well-clamped ...

The mating method will effect the tolerance. I bet you can get a 10% swing just by varying pressure manually.

Your tolerance is impractical and probably unnecessary in a power circuit.

RL

Lots of cheap ferrites saturate at a few hundred Tesla.

John

If you do this a lot I suggest to get a dipmeter and an assortment of precise caps in the 10pf to 1000pF range. Works like this:

Solder cap to winding, no core in there, read resonant frequency, calculate inductance. If it has to be precise hang a good DSO, counter, analyser or whatever near the dipper coil to get the precise frequency.

A dipmeter shows you the width of the resonance and thus roughly the Q.

Considering that's a fair bit higher than any continuous magnetic field produced on earth, it doesn't sound like too much of a limitation. ;-)

clean and well-clamped ...

I agree. I bet the inductance will vary even more at different currents. Usually it is good to have an inductor with a lower inductance at higher currents.

How do you come to this "12%" off figure? Is this some idealized calculation from a handbook or a magnetics designer program? The effective AL value will change depending on the fill of the windings. Since you're doing a flyback converter, your fill will probably be around half, thus, I would expect your inductance to be less than predicted. Clamping pressure will change things also.

BTW, a shorted turn will lower your inductance a whole bunch. Basically turns your inductor in to a piece of wire if the coupling is high enough.

?? Did you mean that previous poster forgot "milli" prefix.?

I thought Tesla was a car ...

I guess so. Saturation of ferrites is typically a couple thousand gauss, which would be a couple hundred mT (1T = 1E4 gauss).