High voltage , high frequency caps in half bridge?

[grizedale]

Hi

I am doing 250W half bridge (offline)

I am trying to save money, and using all these caps as 680nF.

THE peak power is much more than the average.

is this ok

switching frequency is 111KHz.

Also, do you have example schematics of offline half bridge SMPS's of power level a 100W - 500W?

 

[kleanoil]

I think it's workable.

 

[ea.arun]

here is one using LM5035....

**broken link removed**

 

[mtwieg]

here is one using LM5035....

**broken link removed**

This uses 6.8uF caps, ten times what grizedale wants to use. It doesn't really give any advice about the sizing of the capacitors.

I presume that you want to keep the ripple voltage at the capacitor divider node low so that it doesn't have a significant effect on the volt-time product seen by the transformer. Other wise it will affect the open loop transfer function of the converter, and the change will be load dependent. I would suggest doing a mathematical analysis of this effect and just decide how much ripple is tolerable. It might be only %1 of Vin, or it may be %10. I can't really say.

 

[grizedale]

6u8's with th enecessary low dissipation factor, would be huge film caps, and massively expensive, so that is out of the question.

 

[FvM]

Capacitors have an AC current specification, that should be kept.

6u8's with th enecessary low dissipation factor, would be huge film caps.

Increasing the capacitance means, that the AC voltage decreases, thus a higher dissipation factor can be possibly tolerated. But it's still the current rating, that matters.

 

[Orson Cart]

Since the unit is half bridge why not split the bulk input caps instead? This gives more surface area for the bulk caps and much the same cost and you don't need other caps... Regards, Orson Cart.

 

[grizedale]

Hello,

Orson......you have posed an extremely , and i really mean , "extremely" interesting question when you speak of simply splitting (i.e. putting two in series , of twice the single value) the bulk electrolytic input capacitance, and using that as the rail splitting capacitance for the half-bridge primary-side.

But if you see page 4 (top) of the following

**broken link removed**

....it shows the rail-splitting capacitors separately to the bulk capacitors.

-so, simply splitting the bulk electrolytics is not the "usual" way to do a half-bridge, though despite huge trawling searches through the web, ive never found a single article which states why the split capacitors are implemented as separate film capacitors rather than simply splitting the bulk electrolytic up.


Does any reader have a theory as to why the rail splitting capacitors , in half bridge smps, are not often implemented by simply splitting the bulk electrolytic up into two electrolytics in series?

 

[mtwieg]

Theoretically there shouldn't be any problem with using the splitting capacitors as the DC link capacitors. I imagine they're normally separate because it's difficult to find one capacitor that fits the specs for both purposes. For the DC link caps, you mainly want as much capacitance as possible to meet holdup requirements, and enough ripple current rating to work with the current pulses from the AC rectifier. For the bus splitting caps, you care more about high frequency impedance and ripple rating rather than bulk capacitance. It would be difficult (and expensive) to get a high frequency capacitor with enough capacitance to meet normal holdup times. So it's probably cheaper to have different sets of caps.

 

[Orson Cart]

Mtwieg has a point, the bulk caps must be able to handle the HF ripple current as required, this is usually easily possible with reasonably good electrolytics - also the ripple current rating increases with freq (often 30% higher or more) also the ripple current is shared between the caps, we have designed this type of 1/2 bridge for our customers and tens of thoudands have been built and sold without problems - remember you can't use standard peak current mode control though, Regards, Orson Cart.

 

[grizedale]

Orson......so i believe you are imlpying that half-bridge smps's can be built with no film capacitor divider....but just split the bulk electrolytic into two in order to provide the capacitive divider?

mtwieg, -everything you say makes good sense to me, but i disagree that its cheaper to have a bulk capacitive bank of electrolytics and then a capacitive divider made of film capacitors.

Film capacitors of a few uF, and 250VDC rated are extremely expensive and considerably large.

If you do use a film capacitor divider, you find in your layout, that the film caps take up a lot of room, and then the electrolytics end up being some distance from the half-bridge transistors........and since some high-frequency current still does come from the bulk electrolytic bank (even if a film cap divider is used) , you end up with a large high frequency current loop which is obviously unwanted.

I would say that in absolutely all cases, it is cheaper and better to NOT use a film capacitor divider, but simply to split the bulk electrolytic capacitors in to two in series, and use them as the capacitive divider.

-obviously some small , high voltage ceramics can be used across the electrolytics.

-But then my question is, since this debate has revealed not one proper reason for using a film capacitor divider, why do most textbooks and application notes show a film capacitor divider being used with half-bridge smps's?

....it is simply not correct to say that film capacitors are needed to handle the high frequency switching current ripple, because it would be cheaper to simply install electrolytics which are rated to handle this ripple........i.e the electrolytics in the bulk bank which would be split to provide the capacitive divider.

So. does any reader know why film caps are used?

 

[mtwieg]

mtwieg, -everything you say makes good sense to me, but i disagree that its cheaper to have a bulk capacitive bank of electrolytics and then a capacitive divider made of film capacitors.

I never said anything about film capacitors! I just said that electrolytics alone aren't suitable for bus splitting.

-obviously some small , high voltage ceramics can be used across the electrolytics.

Sure, that might work. There will probably need to be something to help keep the very high frequency components of the switching currents confined to small loops to avoid EMI, and ease the burden on the electrolytics.

 

[grizedale]

"I just said that electrolytics alone aren't suitable for bus splitting."

-I cannot understand that.....one simply has to increase the electrolytic rail splitting capacitors until they meet the ripple requirement.....there is surely no earthly reason why electrolytics alone can't do the bus-splitting job?

 

[Orson Cart]

Above 150kHz, electrolytics tend to become more inductive, hence using local film caps in parallel with the electro's - Regards, Orson Cart.

 

[grizedale]

Thanks Orson........though i suppose i could be contrary and ask why , in that case, don't all smps's use film caps for this purpose.........i dont see film caps in Full-Bridge schematics.....nor in flybacks

 

[Orson Cart]

That's because soft switched or resonant full bridge power stages don't need good decoupling like a hard switched PWM power stage does, similarly for flyback - close decoupling is not essential, Regards Orson Cart.

 

[grizedale]

so i presume that hard-switched full-bridge smps' do need these film capacitors, or rather "capacitors with low ESL and low ESR".?

 

[mtwieg]

You're getting into the gory details of design, so you really shouldn't have to generalize these things. I'm sure that some designs would work okay with using electrolytics alone, but at higher switching frequencies you'll want some ceramic or film caps to help suppress EMI and common mode noise which electrolytics can't suppress. But yeah I agree that it is bizarre that many schematics and app notes completely separate the DC link electrolytic caps and the bus splitting film caps. That doesn't make sense to me either.

 

[Orson Cart]

It is definitely true that some power topologies need better decoupling than others, and if a Tx with high leakage is being hard switched, Regards, Orson Cart