mosfets, smps, snubbers, values?

[duck_arse]

I was thinking about snubbing, and looking for a drawn version of the circuit following, when I instead found the circuit following:



I see it has the voltage rating marked on the snubber cap, but many times I see the other version of this diagram without the -33V, without the cap volts noted. so, to the questions:

the cap in the snubber network across the mosfet gets an ohms law-load of the output volts, doesn't it? people fitting 25V ceramics and winding the output up to 200V or more are heading for a trauma, no? how do we pick the snubber values, given an frequency of operation?

and are we snubbing the inductor spikes or the rectifier clangs?

[PRR]

> fitting 25V ceramics and winding the output up to 200V or more are heading for a trauma, no?

No. A 25V ceramic may pop-and-die with very little drama/trauma. Little switchers won't spike-kill large MOSFETs, so lack of snub may go un-noticed.

If the rectifier is perfect (fast), the snub is not needed. Since it never perfectly fast, we like to damp the loose end of the coil so it doesn't whip-snap the MOSFET. However the big IRF are hard to kill with little coils.

[amptramp]

Check out the Unitrode (now part of TI) snubber design:

http://www.circuits.dk/texas-instruments-unitrodeswitching-power-supply-design/

and go to 1993 Snubber Design.  In fact, the whole seminar series is worthwhile reading.

[duck_arse]

thanks, gents. more reading.

[R.G.]

To add to what Paul said:

What needs snubbed are things that cause deposition of big amounts of RF when they turn off. If PCB traces were perfect short circuits, this circuit would not need snubbers, because the step-down inductor is "snubbed" by delivering current into the output filter cap.

What's not perfect about this is that PCB traces have inductance too, so every length of copper may act like an inductor.
The corollary to this is that snubbers need to be separated from the things they snub by zero-length PCB traces, or as near zero length as is possible. In this case I'd put that RC snubber right across the MOSFET. This is not the ideal position for EMI, as it lets the traces ring, but it protects the MOSFET - if, as Paul notes, the MOSFET needs snubbing. Modern power MOSFETs are "avalanche rated", which is a code word for "you can break this thing down with voltage spikes and as long as it doesn't have to eat so much energy it overheats, it's OK".

The better location for the snubber is two places: across the inductor, and/or at the diode. The diode characteristics and snubbing will have major effects on EMI generation. So will the ESR and ESL of the output cap. The output cap itself can ring if you get unlucky with switching frequencies. The diode itself may need snubbing to cut down on EMI.

As a final note, there are single-chip circuits specifically intended for doing what you're doing here - running a flyback up-verter. They're generally 8 pin DIPs, have cycle-by-cycle current protection on the MOSFET, advanced regulation features, and better internal tolerance for the oddities of high frequency switching parasitics as well as soft start.

Have a look at the UCC38xx series of power control ICs at TI.