Half bridge converters: Why two caps?

Two caps act as both the DC voltage eliminator for the winding and also as the energy storage across the DC bus.

Shorter inductive loops to both fets.

John

Ok, that would make sense in apps that need large electrolytics but for a converter running at hundreds of kHz I don't quite see an advantage. With large electrolytics they would each have to have the full DC rating, else it'll be kablouie if the bridge hangs. Somehow that doesn't appear to be a bargain, on the contrary, you could go with one electrolytic and then some small ceramics as couplers for the bridge itself, and save some money.

Ok, but with todays 0805 or 0603 the traces and the transformer leakage inductance win anyhow.

How about startup transient before the single cap charged up. The transformer must be designed to see full DC voltage for a few cycles without saturating. Also the that full voltage will be reflected across the transformer to the secondaries and might either cause problems or require the peak voltage rating to be somewhat higher.

It keeps the high frequency switching current out of the main filter cap. The current could be significant in high power apps.

This means the main cap doesn't need low ESR and low inductance, which makes it expensive.

The two caps in series carry all the switching current and need low ESR. But they can be smaller and cheaper since they run at much higher frequency.

For low power applications, it probably doesn't make much difference.

Regards,

Mike Monett

But at 100kHz, no electrolytics have a very low impedance. With only 1 cap in series with the transformer, one half cycle connects a mosfet across that cap and the transformer, but the other half cycle connects the inductive electrolytic in series with the cap and the transformer. Not very symmetrical. At operating frequency, the electrolytic should not be involved in the current, much. In fact, isolating it with a ferrite bead would be a good idea, to keep the operating frequency current out of it for noise containment.

Doesn't work in either case. If the half-bridge output sits at GND as it usually does when idle and then start you'd have the same issues. At the end of the day the circuit just has to go through a phase where some saturation is inevitable. That's why mine all have a soft-start by means of a current limiter.

Not really. If you have just one cap in series the other end of the transformer goes to GND. The VCC side of the bridge needs to be bypassed by ceramics anyhow.

Ok, but any engineer in his right mind will bypass the bridge supply with one or more good ceramics. There is (usually) a lot of other stuff hanging on it that requires a clean rail.

I tried it out a few times and there was no difference at all, except for the extra cost for the 2nd cap which I saved ;-)

If input noise compliance is not needed, as input is from a PFC output bulk cap, go with the one cap configuration. If the input must be filtered to compliance standards, then the two cap configuration is the best choice. It will increase the input ripple current frequency and reduce the peak current by a factor of 2. This will reduce your input filter requirements by

Harry

Conventionally, an 'idling' half bridge is off - high-z. The appearance of the power rail produces 1/2 the supply rail at the junction of the two-cap circuit, through KCL, with no control or power circuit involvement.

A single cap obtains charge equalization only through the active bridge and magnetic structure being driven, not always with benign start-up effects.

Originally, the availability of rectified-line-rated film capacitors with ripple current ratings in the ampere region was limited, and the cost of larger parts was prohibitive. To reduce cost at lower power levels, a single cap was often used - sometimes coupled to he center-tap of the electrolytics in the input voltage doubler, when present.

Current through either configuration must still pass between the power rails, through whatever decoupling is present - typically the bulk electrolytics and a (usually overwhelmed) film cap.

The circuit you use will either work or not, depending on the circumstances, and the care you take in their anticipation. Until you appreciate the full ramifications of a component's function in the circuit, prudence would suggest that simple accuracy is required, in attempting to copy someone else's design.

RL

The ones I came across weren't that way. They always leaned to one side when idle. The only time they'd be coasting was during transitions (dead time, to avoid cross conduction).

Ok, early availability issues, that makes sense.

I never copy circuits, it all starts with a blank sheet of paper. It's almost unavoidable to touch core saturation here or there, even with the two caps (they'll never be 100% identical). But that's not all, there is also the issue of charging up the filter cap on the other side. Even with a smoothing inductor there you'll exceed nominal current during start-up. So I always have a soft-start in my bridges. Basically an input current limiter that pulls the bridge to the side. This also avoid most of the sparks when someone drops a wrench onto the board that's connected.

Looks good, I guess. Symmetry is appealing. We all know if there's C between the rails and C to that node, it doesn't matter whether you use one, two or a million, it's still a bypassed node. Setting initial voltage isn't a bad point though.

Every half bridge I've seen (computer PSU) uses a 1uF, maybe 2.2uF, film cap to the CT between the two electrolytics. This solves coupling and initial voltage while using only one capacitor, but doesn't solve bypass. The supply 'lytics are in series with the bridge halves, so they must carry signal current, and it's absurd to state that they don't (cf. John Popelish above). It's possible to bypass them with more stout capacitors, like non-L films and ceramics and stuff, but nobody does it. Well, maybe people who make quality stuff... computer PSUs are obviously no quality reference. :-)

Tim

each half cycle uses one of those 2 capacitors, if there was one missing the high frequency part of the one half of the switching current would be taking a longer route via the electrolytic with increased impedance. this is bad for rfi and voltage acros the fets.

you could have a smaller hf bypass cap, or a cap across the electrolytic but then youve got 2 caps still, cheaper to have 2 the same than 1 big 1 small.

the dc offset is an issue depending on the size of the capacitors but could be dealt with by carefull soft start.

many PC smps ive seen use 2 190v electrolytics in series with a full bridge for 240vac operation and a voltage doubler for 110 volt operation. although the voltage doubler isnt seen in recent ones ive looked at. you still need hf bypass capacitors but not as big.

the most recent one ive seen has a PFC module. ive recently built up a small collection of PSU with faulty capacitors. I might make something interesting out of the bits one day im not sure what, the plasma cutter I worked on once was cool.

Colin.

Yes, this may go back all the way to the olden days. Splitting electrolytics in a half-bridge and hoping the voltages divvy up correctly is playing with fire though.

I probably was too stuck in my world of DC/DC converters where there really aren't any electrolytics.