Full-Bridge SMPS vs Phase-shift-Full-Bridge SMPS?

[grizedale]

Hello,

We need to do a 700W offline converter.

We are decidning between phase shift full bridge and plain full bridge.

It appears that the phase shift version needs no additional hardware and gives zero switch-on switching losses.

So my question is, since the phase shift version needs no additional hardware, why does anybody ever do a plain full bridge?

Why doesnt everybody always do a phase shift full bridge?

 

[mtwieg]

Major reason is that it's not well understood by most people.
Also I'm pretty sure you need a specialized controller or some extra external circuitry to properly achieve ZVT. You need to insert some a relatively large dead time into each half bridge, which most controller ICs can't do (though doing it externally is quite simple). I've heard that having a MCU controller is excellent for phase shifted bridges since the dead time can be appropriately adjusted to give good performance over a wide line/load range.
Getting repeatable behavior out of a design may require tight control of the transformer leakage, and the FET capacitance, requiring those to be implemented as discrete components (more cost).
Also the side effect of resonant converters is that you lose some effective input voltage (or to put it another way, you're effective "duty cycle" can can't go to %100).

But yeah the phase shifted bridge is nice if you grasp it well, and actually benefit significantly from its perks (mainly you're operating at high frequency).

 

[grizedale]

is it really that difficult?

surely all that happens is that instead of just turning two off at the same time.......one simply turns just the low side fet off..........the current then goes through the diode of the opposite side high side fet.......so it comes on with ZVS....then one simply turns off the due-to-go-ff high side fet, and hay presto the current turns on the low side fets diode, , you then simply turn that fet on and it comes on with zvs.

thats remarkably simple, so howcome we believe people cant grasp it?

 

[Orson Cart]

The phase shifted full bridge will give much lower CM-conducted RFI when done properly than a hard switched H bridge.

 

[grizedale]

Well,
-there are controllers readily available for phase shift full-bridge, and they are not really any more expensive than the normal bridge controllers.

-also, i think we all agree that there is nothing difficult about the principle of operation.

-and we all agree that they are more or less superior in every way, so why would anyone bother with a plain full-bridge when the phase shift full-bridge is superior and costs just the same?

-The duty cycle restriction is minor and wouldnt affect most converters.

-The tight spec on drain capacitance and leakage surely is not relevant, as there will still be only a short deadtime which accounts for all situations.

 

[mtwieg]

is it really that difficult?

surely all that happens is that instead of just turning two off at the same time.......one simply turns just the low side fet off..........the current then goes through the diode of the opposite side high side fet.......so it comes on with ZVS....then one simply turns off the due-to-go-ff high side fet, and hay presto the current turns on the low side fets diode, , you then simply turn that fet on and it comes on with zvs.

thats remarkably simple, so howcome we believe people cant grasp it?

It sounds simple until you realize that if the dead time is too high, the diode conduction losses cause more overall losses and suddenly you lose the main benefit of ZVS. The optimal dead time to achieve ZVS while minimizing diode conduction losses is line and load dependent, so it is difficult to maintain optimal performance under a broad operating range (hence why it helps to have a MCU as a controller). Even so, it's normal to add schottky diodes to each FET, which increases cost a bit.

A lot of engineers just don't care enough about an extra few percent in efficiency enough to put the necessary thought and effort and cost to obtain it.

 

[grizedale]

mtwieg, are you talking secondary diodes, or diodes parallel to the primary FETs?

 

[mtwieg]

Parallel to the primary FETs. They're not necessary if the dead time is properly controlled, but that usually requires an MCU.

 

[Orson Cart]

You get the diodes for free when using mosfets in the H bridge - they are quite necessary for when you turn the converter OFF at full power - the Tx pri leakage current must go somewhere and without diodes across igbt's they will reverse bias conduct which may lead to a bang when the Tx energy falls to zero and full rail volts are reapplied to that bridge leg.

 

[mtwieg]

The body diodes of MOSFETs won't be schottkys though (unless they're specially built to have schottkys). The lower Vf helps minimize conduction time during the dead time (slightly).