Switching Mosfet overshoot

[Claude Abraham]

originally posted by tennisballq
The big problem was my transformer core. I was told that Iron Powder would work, but I don't think the permeability is great enough at the frequency I am running, and it lets too much current through. I should actually do the math BEFORE I start, what a novel idea. Anyway, I found that I had blown one of the mosfets (probably due to this). Replacing it put the ringing in its place! There is still a tiny spike on the rising edge, but no huge ocillation as before.

Ok, now I've put it all together and everything works fine....except that I have +/-60V rails now instead of the intended 30V. Looking at the primary, I have 23V pek square waves on each primary. Not sure what went wrong there. On the primary I did 4 turns, center tap, 4 turns, the secondary was 10 turns, center tap 10 turns, and I only have a 12V supply. I'm wondering if I made some minor mistake and it's acting as an auto-transformer. Back to winding again...

I would definitely perform the magnetics computations first, and then wind the transformer. I would use ferrite core material. As far as the 23 V on each primary winding goes, a push-pull converter topology places a voltage stress on each MOSFET drain equal to twice the input supply voltage. About 24 V is what I would expect when the input is 12 V. As far as winding a transformer goes, minimum leakage inductance is mandatory. The biggest problem with custom transformers for SMPS is leakage inductance. Also, use a low leakage inductance core geometry, having a wide window and shallow height. I would keep each winding at one layer Do not put the primary and secondary on the same layer. Each winding should span the full breadth of the window, leaving a small margin on each end, and this way magnetic coupling is maximized. I would use bifilar windings for both primary and secondary, making the center tap connections externally. Each half of the C.T primary should be vertically aligned (directly above or below) its corresponding C.T. secondary half. At 100 kHz switching frequency, I would use wire no thicker than AWG 26. At 50 kHz, up to AWG 23 is ok. Parallelling conductors (if bifilar AWG 26 isn't quite enough, use quadfilar, or hexfilar, etc.) increases current capability, while keeping skin effect under control. Also, multifilar conductors helps fill the window width when the number of turns is small.

Last, you are using an output filter inductor, aren't you? Also, you are using feedback, correct? Otherwise, you will get an unregulated output. For anything more than a few watts, I avoid inductorless circuits. Without an inductor, ripple current on the output caps is very large, making large sized caps necessary. Also, noise is much worse without an inductor. As far as the overvoltage output, my guess would be that under no load a substantial secondary leakage inductance could over charge the output caps. Hard to tell from here. Best regards.

[tennisballg]

Wow, thanks...very detailed. I know I should have figured out which core to use first, but I was told that the powder would work ,and didn't question it! Anyway, thanks very much guys, learning the stuff in textbooks isn't nearly as good as experience.

[Circlotron]

Originally posted by tennisballg
I was told that the powder would work

Probably would have worked for a flyback converter, but not terribly well for a forward converter that you have done.

/Circlotron - not scared of winding switcher trannies either.

[tennisballg]

This is beginning to drive me crazy!. Don't get me wrong, I won't give up, but I am so close to having it done but for one problem. Everything works, I have a beautifully straight DC line, one positive, one negative, but no matter what I have done, one rail is always significantly higher than the other. Not exactly double, but about that. I've gone through, rewound the transformer a couple of times, checked the orientations of all the capacitors/diodes, and can't find anything out of the ordinary. I mean, it's such a simple principle to rectify a wave. The nearest I can figure is that I have some little error somewhere that is causing a voltage doubler with my caps and diodes. I guess the next step is to remove the caps and see what happens...

[jan.didden]

....schematic diagram, perhaps?

Jan Didden

[Sei-Young Jang]

switching freq., Input voltage, Topology, magnetics data such as primary inductance,number of turns, and schematic. SMPS is not so simple as some think.

[Eva]

Have you tried to put some symmetrical load on the output?

[tennisballg]

I suppose I should try loading it before I undo my work. I didn't mean that the SMPS would be simple, I meant that rectification should be.

[subwo1]

Actually, for half-bridge or push-pull or full bridge for those into that topology, the rectification part is important because it may be helpful to draw power off the transformer during each waveform cycle. So I recommend avoiding a half-wave rectifier in these cases. For flyback topology, a half wave rectifier may be useful. Also, be sure to include a choke between the rectifiers and the filter capacitor. It took me a long time to discover its importance.

[Symx]

Check out this post from earlier in a previous Topic.
As I,ve pointed out you may need to slow down the swithing a little to overcome this problem. SMPS problem