SS rectifier with parallel caps

[FREYES_07]

Hello all,

Been looking into ss rectifiers, and found the following topology in master blencowe's book

But I'm wondering why doesn't (or does?) the current flow as shown in the poorly drawn schematic attached?

Wisdom will be well appreciated

[GAStan]

How a full wave rectifier works

I cannot explain it any better than it is here.

[martin manning]

I'm wondering why doesn't (or does?) the current flow as shown in the poorly drawn schematic attached?

It does, but the capacitive reactance of the typical 10n used in that way is 265k ohms, two in series 530k.

[GAStan]

Apologies FREYES_07, I didn't fully understand the question until Martin answered it.

[R.G.]

Yep, +1 to Martin!

This brings up in my mind the question of why someone would put those capacitors there. I can think of two reasons off-hand. One is to "snub" the diode turn-off transient for non-fast/soft turn off diode, and the other is to force voltage sharing on the diodes during times of rapid voltage changes.

Snubbing ordinary recovery diodes has a long history, and is a good idea for plain-vanilla diodes (if snubbing the transformer itself isn't done). My guess is that a "snubber" capacitor across one diode got into designs, and then someone decided that two diodes in series was needed for reverse voltage reasons. One capacitor across both series diodes works the same way as one cap across one rectifier as far as snubbing the turn off is concerned. But if the diodes in series don't have matching turn-off times, the faster of the two turns off before the other one does and is left holding off all the reverse voltage until its series buddy diode turns off too. So the faster turn-off diode can fail in reverse breakdown over time from the transients. We're playing in the sub-microsecond arena here.

Two caps in series helps force the voltages to be equal on sudden transients and acts like a voltage divider to prevent the faster turn-off device from being over-volted if that was what was going on.

A single cap across one or two diodes is not the best snubber to prevent 120Hz noise from diode turn-off, as in some cases the cap can tune the wires' inductances into an RF resonator and make the perceived RF-detected hum worse. Rare, but it happens. Inserting a resistor of some sort in series or parallel with the caps helps by adding damping to any unfortunate RF resonances.

Notice that the snubbing function of the caps is not low impedance. The 200K+ values at power line frequencies drop by 10,000 to 1M at high speed transients - like rectifier turn-off.

[martin manning]

My understanding from something i read a while ago is that UF-type diodes do not need snubber caps, but I'm not sure if the voltage balancing effect would still be desirable.

[trobbins]

if the diodes in series don't have matching turn-off times, the faster of the two turns off before the other one does and is left holding off all the reverse voltage until its series buddy diode turns off too. So the faster turn-off diode can fail in reverse breakdown over time from the transients.

It would certainly need a significant transient to couple through the power transformer, and perhaps be time-aligned with worst-case voltage waveform position, and then for one diode to enter avalanche conditions such that its avalanche induced increase in current also flows through the other diode whilst it's transitioning states but perhaps has a lower voltage across it (but still has voltage headroom to stay below avalanche conditions). Not sure if I've seen that mechanism exposed in a detailed way.

[R.G.]

My understanding from something i read a while ago is that UF-type diodes do not need snubber caps, but I'm not sure if the voltage balancing effect would still be desirable.

The need for snubbing comes from the semiconductor physics of ordinary rectifiers. Rectifiers have to have large-area junctions for the currents, and largish depth depletion regions to hold off high voltages. That volume gets filled with charge carriers when the rectifier is conducting, and has to all be drained out before the newly-empty depletion region/junction will support reverse voltage. When the depletion region empties, the junction will abruptly stop conducting. Abruptly, much like slamming a door. It's a fast, HARD turn-off. UF type diodes ar both fast (don't have the same got-to-empty-it-all effect) and soft (slowly recovers voltage-supporting ability, not all at once) so they don't make the brick-wall turn offs that couple out as sub-microsecond transients. They don't excite ringing and RF nearly as much because the slammed-door doesn't happen.

If you really need two or more diodes in series for the voltage, you might still need balancing resistors to swamp out differences in reverse leakage. If you have, say, 1500V reverse voltages and can only get 800V UF diodes, using balancing resistors ensures that even if one diode has twice the leakage of the other, both diodes see only about half the reverse voltage, instead of 1/3 and 2/3 because of the leakage. This is much less of a problem today than it was back in the bad-old-semiconductor days, as leakage is massively lower in all types now. It's for sure that balancing resistors will not hurt as long as they don't themselves cause too much reverse current.

It would certainly need a significant transient to couple through the power transformer, and perhaps be time-aligned with worst-case voltage waveform position, and then for one diode to enter avalanche conditions such that its avalanche induced increase in current also flows through the other diode whilst it's transitioning states but perhaps has a lower voltage across it (but still has voltage headroom to stay below avalanche conditions). Not sure if I've seen that mechanism exposed in a detailed way.

Yes, it does take a chain of events. I wasn't thinking of a power line transient coming through the transformer so much as the slam-and-ring of one of the diodes in the rectifiers slamming off and letting the wire inductances ring. As much as anything else, I was mentally exploring what could happen with the rectifiers themselves causing transients. That's automatically aligned with the worst-case voltage position, or nearly so since the diodes all act right at the peaks of the waveforms.

[martin manning]

Thanks for the explanation!