[SOLVED] Design "Soft switching" into a MOSFET full bridge.

[Reactor89]

My application:



This is an idealize full bridge switch mode DC-DC boost power supply as I know it (with a MOSFET full bridge drive and full bridge rectifier). The power supply output is 300 volts at 1 amp nominal and 2 amps peak output. Power supply input is 12 volts at a necessary amount of current.

Under this configuration the MOSFET full bridge (Q1 - Q4) would "hard switching" the transformer (T1) wasting energy and creating extra heat in the the process.

I'd prefer that the MOSFET full bridge would "soft switch" the transformer and not "fight" with the transformer and create extra MOSFET heat so to speak.

It's at this point that zero voltage switching (ZVS) and zero current switching (ZCS) should be brought up as possible soft switching methods.

From what I can tell, a ZVS version of the same power supply would look like this:



I have no proof that this ZVS is what I need to create the soft switching I want and I don't have any concrete math for calculating the values of L2 and C3 to C6 in practice. Yet any information about ZVS in this application is appreciated.

Any suggestions about how to design soft switching into this power supply (Using ZVS, ZVC, or other method) and formals calculating the necessary component values is greatly appreciated.

Thanks in advance.

 

[Warpspeed]

With only twelve volts dc input, by far the biggest problem to overcome will be conduction loss, not switching loss. Switching loss will be negligible in comparison at any reasonably high power level.

Deliberately adding extra conduction loss in an effort to further reduce an already negligible switching loss is completely counterproductive.

 

[dick_freebird]

Soft switching attempts to minimize the transient conduction
loss in the low side switch. Although this may be allocated
to "switching losses" because it is periodic and transient, it
is an ohmic, not capacitive-shuttle, Joule slug.

Basic idea being that if you delay the low side switch
turnon until the inductor current has swung the node,
it's (VIN*Coss_effective) a freebie. You don't get to lose
the Vgs*Ciss_effective however, just reposition it. The
Miller charge, that varies.

I do not agree or disagree about the value of minimizing it
by effort / elaborateness; that wants close-in analysis, a
quantification of what performance stands to gain. And
this also has to comprehend downsides such as, can you
run at your minimum and/or maximum desired pulse width
if the falling edge is allowed to be leisurely and variable
(depending on inductor current for dV/dt)? I've only walked
away from soft switching after looking at the troubles in
making it feedback-controlled (although this is a key part
of some successful DC-DC outfits' designs, clever enough).
But is it worth the pain, to be so clever? Comes down to
cases.

I'm not convinced, as OP states, that soft switching is
about the transformer at all. More about the totem pole
output nodes and getting them to the right place at the
right price. Transformer is just the meat in the middle of
the sandwich, where you're trying not to burn the toast.

 

[mtwieg]

You want to review literature on phase shifted full bridge converters, which are designed to reduce switching losses by achieving ZVS. Also look at LLC converters, which can be operated with ZVS or ZCS (but generally ZVS is preferred). There is tons of literature about both available online, though the LLC gets somewhat messy in the math department.

Keep in mind that improved power efficiency isn't the only benefit of soft switching converters. At high power levels, meeting EMC restrictions is often a great challenge, and soft switching converters can greatly reduce EMC relative to a hard switched converter at the same power and frequency.

 

[Reactor89]

Thank you for all the helpful replies Warpspeed, dick_freebird, and mtwieg.

Warpspeed, thank you from bringing me back to reality on this design. ZVS is primarily for reducing current ripples and transient voltages created with load switching. Using a low voltage, like 12v, guarantees high currents with lots of noise and conduction losses. Conduction losses that won't significantly go down by simply filtering the switching circuit.

This being said my design priorities for this power supply are as follows; use high quality (expensive) parts, create reasonable efficiency, and have minimal concern for electromagnetic radiation. I.E. I'm building a power supply for trained consumers that really need such a power supply.

To this end, the MOSFET I want to use for the full bridge is the Vishay Siliconix SUP90N06-6m0P N-Channel MOSFET (https://www.vishay.com/docs/69536/sup90n06.pdf). With an RDS(on) round 6 mohm max, my conduction loss with will be somewhere in the running range of 5 to 30 watts depending on the power being requested (in this case 240 watts and 600 watts respectively).

Now that I put some numbers to the design is it still a responsible assumption that the conduction losses are drastically larger had the switching loss?

My current plan is to, not be clever (at least to start), implement hard switching at the transformer's most efficient frequency, and create reasonable power supply. Then do more reading on phase shifted full bridge and LLC converters designs if need to build lower power converters in the future.

 

[treez]

Re: Design "Soft switching" into a MOSFET full bridge.

Please do LLC converter, because that will be the best way to give you lowest voltage stress on your output diodes...this is important for you as your output voltage is high. The PSFB converter is good ,but it will give you the problem of ringing on the output diodes which means you will need higher voltage output diodes there.
Here I send you docs on LLC.

- - - Updated - - -

tHE MATHS OF LLC IS NOT TOO HARD AS LONG AS YOU DONT HAVE TO DO THE F.H.A ANALYSIS BY YOURSELF,,,AND OF COURSE, YOU DONT HAVE TO , BECAUSE IT IS ALREADY DONE, (SOrry about capitals). iT IS PROVIDED HERE IN THE ZIP FILE....AN-4151 by fairchildsemi.com


The attached shows "maths required for SMPS design"...and this is generally all you have to know, even for LLC...since as you know, the hardest maths of SMPS is the PWM Switch model for feedback loop compensation (or state space method)...but you don't necessarily need to do that, because there are standard form equations provided, and anyway, you can check stability with a frequency analyser, and the "measurement always wins"...no matter what anyone calculated......Even if a professor calculates stability..we still believe the frequency analyser for exact value of gain and phase margin. We apologise to the professor, but we say, "sorry but we believe our measurement of gain and phase margin more than your calculation of it...measurement is king"

 

[Warpspeed]

With an RDS(on) round 6 mohm max, my conduction loss with will be somewhere in the running range of 5 to 30 watts depending on the power being requested (in this case 240 watts and 600 watts respectively).

Don't forget that the mosfet is not the only resistance in the circuit.
Conduction losses include everything the inverter input current has to flow through.
With the transformer high current primary, skin effect can be a problem in that the real working resistance of the winding can be much higher, sometimes several times higher than the dc resistance.

Another point to ponder, is that all these conduction losses rise at the rate of current squared.
The most efficient way to transfer power with minimum conduction loss apart from pure dc, is with a 50% 50% duty cycle square wave current. Where the peak current is no higher than the average. Peak, rms, and average are all the same.

Anything else, including sine waves, will have a higher rms than average.
Your inverter transferred power (real watts) will still be the average dc input current multiplied by the dc input voltage.
Conduction losses will be rms input current multiplied by voltage which is highly dependant on the current waveform.

Any current switching waveform that has a lower duty cycle, with higher peaks with a lower average will be more (conduction) lossy than a square wave for the same transmitted power.

Its the single reason why most very high power switching power supplies use forward converters with (almost) square wave currents, especially where the input is a low voltage at relatively high current.

Even with perfect sine waves the conduction loss will be 11% higher than with perfect square waves where the same power goes through the same transformer at the same voltage and frequency. Rms and average are not the same thing.

 

[treez]

Its the single reason why most very high power switching power supplies use forward converters with (almost) square wave currents, especially where the input is a low voltage at relatively high current.

I see what you mean, but in this case the Nsec will be a fair bit higher than Npri, and thus also the L(leak)sec will be higher than L(leak)pri.....so OP could see much ringing across secondary diodes which OP will need to snub out, and even then will likely need higher voltage rated output diodes than OP could use with an LLC

 

[Warpspeed]

Higher voltage secondaries are always a problem.
Stay capacitance and lower resulting self resonant frequency can become a big reactive load during switching.
Winding technique is important and sourcing suitable fast high voltage diodes more difficult.

As a high voltage secondary will almost always require at least two layers on the transformer (with the primary in between) it is entirely practical to rectify and filter each half of the high voltage secondary individually. Then connect the dc outputs in series.

 

[Reactor89]

Don't forget that the mosfet is not the only resistance in the circuit.

Absolutely and further data to say switch loss isn't appreciable compared to all the other losses in the circuit. A boost power supply like this will be lossy compared to buck power supplys and there is only so much you can do about this.

you will need higher voltage output diodes there.

My chosen rectifier diode for this power supply is Vishay VS-10ETF06FPPBF, the 600V version (**broken link removed**). I'm hoping this diode will handle voltage ringing in the full bridge rectifier circuit. I could also try to snub said ringing on all four diodes like this:



Applying filters around the rectifier, on the lower current side of the power supply, is far more appealing to me than trying to filter the high current front end. I would also add resistors in series with the capacitor if the impedance of the filter is too low.

Again, thank you for the input everyone, much appreciated.

 

[treez]

Regarding LLC converter.....
Also, the leakage inductance in your transformer is no longer an ‘enemy’ to you, and so you can scrub “Interleave winding” if you wish, In fact, you can , if you wish, use a split section bobbin (sectioned bobbin) and put the primary coil in one section and the secondary in the other. This can give you the leakage inductance term that you want for the resonant inductor (or at least part of it). You can either use a specially molded sectioned bobbin or just use margin tape to make the “section wall” yourself.

If you have high current, in a high power LLC, and you are confined to using round-cross-section enamelled copper wire, then you will tend to find that using a sectioned bobbin makes it far easier for you to achieve lower winding losses. This is because you are not having to make up turns out of multiple parallel strands of round enamelled copper wire so as to “fill the bobbin length”. –That is actually a difficult point to explain, but if you go about the winding process, making out the winding manufacturing document etc, you will see what I mean.


So, if you do LLC , and you are using round cross section ECW, then you will find the LLC in split section mode allows you to get the lowest possible conduction loss........if you try it, you will see why. I think you have to go through the motions to see it.

And also, as you know, an LLC converter does not always require high primary circulating current, that is, low transformer magnetising inductance. Sometimes yes, but not always.

Regarding the full bridge or PSFB with almost square profile primary current, I think that could suffer from noise issues, and difficulties in control.