Two-switch: Flyback or Forward?

[kathmandu]

Hello,

I want to upgrade one of my 48VDC > 24VDC (100W) converters by using a dual-switch topology.

For now I'm using a single switch flyback circuit. After reading some two-switch related literature, looks like there's no need for that extra (reset) winding anymore (forward topology).

I would like to use a forward topology as it doesn't need a gapped core and the core itself could be smaller.

Is there any (other) downside I'm not aware of?

 

[treez]

The limitation to 50% max duty cycle...however, the hyperTFS from power integrations has a special way to get max duty above this.

 

[kathmandu]

Oh, I see. It just doesn't need a separate reset winding anymore but there's still a reset operation (through those two diodes).

So, for my design, do you recommend the dual-switch flyback topology? I might use the same transformer (aka coupled inductors) from the old circuit?

 

[FlapJack]

Katmandu, i think you have a good plan. I would stick with 50% duty or less and use slope compensation. Then you have a multitude of chip choices from all semi manufactures.

 

[treez]

as you know, you shouldnt really need slope compensation if you have 50% or less duty cycle and in current mode. The hypertfs chip is voltage mode so for that you wouldnt need slope compensation ANYWAy.
2 switch flyback no good....it has only one output diode for the high current whereas two tran forward shares the current between the 2 diodes.
You just need a high side fet drive but that can be done with a bootstrap high side drive chip and many posts in the last 3 months here have discussed this.

- - - Updated - - -

Warpspeed has kindly discussed use of bootstrap high side drive chip for two tran forwards in either this thread..
https://www.edaboard.com/threads/348884/
..or another thread round about the same time...i forget which one.

 

[FlapJack]

Treez, you do the same thing i do. Every time i go for the a key i hit the caps locks!

 

[kathmandu]

2 switch flyback no good....it has only one output diode for the high current whereas two tran forward shares the current between the 2 diodes.

I have a very powerful Schottky diode (60Amp, TO-247 package) thus there should be no problems with single diode output rectification.

You just need a high side fet drive but that can be done with a bootstrap high side drive chip

Yes, but you need a high side driver for both dual-switch topologies (forward/flyback) anyway. And there's plenty of low/high side drivers too.

Katmandu, i think you have a good plan. I would stick with 50% duty or less and use slope compensation. Then you have a multitude of chip choices from all semi manufactures.

Thanks for your support. So you recommend the forward topology or did I get it wrong?

Btw, I don't need a very accurate output. I'm going to use this circuit to supply 24VDC to a lot of other DC converters (based on LM2596T mainly) which don't accept an input voltage greater than 37VDC.

 

[treez]

yes sorry about that, this keyboard i have here is shocking.
48->24vdc could be via (synchronous) buck , so i hope you do need the isolation

 

[kathmandu]

48->24vdc could be via (synchronous) buck , so i hope you do need the isolation

I've talked about it in another thread: if a Mosfet fails short in a buck converter, you got VIN at its output which (in my case) will blow away all those smaller DC converters.

I don't need isolation but this topology it's much safer.

 

[treez]

In that case, go for a sepic, that gives you intrinsic short circuit protection , because the series capacitor gives you DC blocking.
Also, even with the buck ....if a buck fet suddenly died short cct, then you would just need a divider->comprator-> disconnect fet.
..And have a TVS downstream of the disconnect fet so that if the fet didn’t switch out straight away then the TVS catches it for those few 10’s of microseconds.

- - - Updated - - -

SEPIC you design as if a 1:1 flyback, but dont use RCD clamp....the diode and fet currents are same as 1:1 flyback...remember size the c(sepic) so it has less than 4% ripple voltage.
Put an RC snubber across the sepic cap.
Sepic cap sees vin across it.....but rings up at startup etc so de-rate it.
If coupled then you're done, if uncoupled, then double the inductor values from the coupled version.

Keep crossover away from the resonant ring frequency of c(sepic) and the inductance. (obviously the "inductance" is only the leakage if its an coupled version).

Also, the mentioned ringing frequency should be lower than the switching frequency, otherwise you just make a wild oscillator.

 

[kathmandu]

Thanks for the details, @treez, but I'll stick with the flyback topology (I guess it's fail-proof enough).

Furthermore, as you've mentioned, I better use a mechanical relay (not a FET) to quicker disconnect the output if an overvoltage occurs (though for a flyback topology, you only get an overvoltage when the output feedback network fails open).

 

[treez]

-or as you know you can look at the bias winding to sense overvoltage,

 

[Warpspeed]

I would like to use a forward topology as it doesn't need a gapped core and the core itself could be smaller.

Is there any (other) downside I'm not aware of?

While its true that a forward topology does not need an air gap, the core will not be any smaller.
The primary still needs the same number of turns on the same sized core for either forward or flyback.

The only real down side of forward is that there must be a choke inductor in the output after the rectifier for it to work properly.

Flybacks work straight into an output capacitor, and the output voltage can be regulated by adjusting the duty cycle, and this alters the amount of flyback energy pumped per cycle.

Forwards have a fixed ratio in the transformer, and the output voltage will always have the same peak value, regardless of duty cycle. If you feed that straight into an output capacitor, it will always charge up to the voltage peak (at light load) and there will be no way of regulating the output voltage.

A series choke averages the output allowing the ability for duty cycle to change the output voltage. An additional function of the choke is to limit the peak inrush current into your switching mosfet at turn on. The choke is not a luxury, it is a vital part of any forward topology.

 

[kathmandu]

Many thanks for this comprehensive insight! That's it, I'll stay with the (dual-switch) flyback topology.

Is there any problem if I'm going to use the same xformer (coupled inductors) from the former one-switch flyback design?

 

[treez]

The dual switch flyback allows you to limit the off state voltage on the fets, to less than with a single switch flyback, but your vin is only 48vdc, so i am not sure if its really worth doing twoswitch flyback for yourself..what is your reflected voltage to the primary?

 

[Warpspeed]

Many thanks for this comprehensive insight! That's it, I'll stay with the (dual-switch) flyback topology.

Is there any problem if I'm going to use the same xformer (coupled inductors) from the former one-switch flyback design?

The flyback transformer you already have should work fine with the dual switch flyback topology.

 

[kathmandu]

The dual switch flyback allows you to limit the off state voltage on the fets, to less than with a single switch flyback, but your vin is only 48vdc, so i am not sure if its really worth doing twoswitch flyback for yourself..what is your reflected voltage to the primary?

My Mosfets are 150V rated but I really hate not to be able to clamp the voltage across their D-S. And snubbers are the next worst things I dislike.

@Warpspeed: thanks for your confirmation.

 

[treez]

ok but by the way a coupled inductor for a sepic does not have to be interleave wound....the leakage doesnt matter much at all, as long as you mind the aforementioned resonant frequency.