Phase shift full bridge converter and destruction of FETs?

[treez]

https://www.infineon.com/dgdl/Infin...D.pdf?fileId=db3a30433f9a93b7013f9f582ddb1fd9

(ZVS Phase shift full bridge CFD2 optimized design)

The above app note, on Phase shift full bridge converters, at the bottom of page 18, says…………….

“For safety reasons the delay time must be adjusted at the minimum load at which the synchronous rectification is activated. Otherwise there is a risk of destruction of the parts when decreasing the load due to an overlap of VGS and VDS”

Why should the FETs be destroyed by the overlap of VGS and VDS?…..I mean, overlap of VGS and VDS occurs in every single hard-switched SMPS in the world, and it doesn’t result in “destruction” of the FETs………………..so why in a phase shift full bridge converter should this overlap result in destruction?

 

[D.A.(Tony)Stewart]

This is not like every other SMPS. There are separate synchronous timing delays from the full bridge primary to the half-bridge secondary,

If the secondary Vgs (on E) is applied when Vds secondary is high rather than low then the dynamic Turn-on losses could be destructive.

 

[treez]

Thanks, ultimatelty, we are trying to find out if a "Non-zero switching phase shift full bridge converter" is any worse than a plain full bridge converter.
Our vin=380VDC.
Vout = 250-400VDC (battery)
Pout = 1.75KW

 

[D.A.(Tony)Stewart]

I like;
1) absence of MIller plateau on Vgs turn-on which causes dynamic loss, especially significant for efficiency improvement over a wide range of loads
2) total losses always< 5% from 25~100%Pmax
3) absence of ringing on primary current and high power factor.

 

[treez]

In relation to this, I believe that the PSFB (Phase shift full bridge converter) is simply not suitable for high output voltage applications, as is confirmed by Dr Hongmei Wan...

Please confirm that the Phase Shift Full Bridge (PSFB) SMPS is not suitable for 7KW Public Hybrid Electric Vehicle battery chargers?

Spec of PHEV charger
Vin = 90-265VAC (single phase)
Vout = 250V-400VDC (7KW)

The high output voltage of 400V, coupled with the much higher output diode stress that occurs with a PSFB means that the PSFB is not suitable.

The following article confirms the unsuitability of the PSFB….

“High Efficiency DC-DC Converter for EV
Battery Charger Using Hybrid Resonant and
PWM Technique”
By Hongmei Wan…………

https://scholar.lib.vt.edu/theses/available/etd-05072012-141855/unrestricted/Wan_HM_T_2012.pdf


This document confirms my postulate….also, Dr Wan proposes that for this application, the way forward is a “hybrid” converter, which comprises both an LLC converter and a PSFB converter, together with a non-dissipative diode snubber network.

Dr Wan proposes this, because as in his report, the PSFB by itself is not suitable for this application.


The alternative to Mr Wan’s complex hybrid converter is just simply to parallel multiple hard-switched Full Bridge converters….which seems like the sensible way forward in truth, do you agree with all this?
.......................................................................................................................................
The absolute *golden* question concerning Phase shift full bridge (PSFB) converters, is that if the dead time and leakage inductance is improperly sized such that zero voltage switching does not occur, then are they more or less efficient than a plain, hard-switched Full Bridge? No-one addresses this point in any document anywhere. I suspect that the PSFB converter is less efficient than the plain full bridge if not “tuned” properly.

 

[mtwieg]

Sure if you look at random MS (not doctorate) theses, then you'll find novel stuff which outperforms the state of the art. Go ahead and surf IEEE and you'll find hundreds of manuscripts describing converters with 98%. But even so, the state of art persists, mainly because the bleeding edge stuff turns out to not be cost effective.

In any case, I don't see why you would need very high efficiency over a large load range for battery charging. Operating in low frequency bursts at higher power solves that issue.

 

[treez]

Thanks, yes i agree on the burst thing.

But supposing that this was done with 4 paralleled CCM boost PFC's.....into 4 paralleled plain full bridge converters...is there any reason why that wouldn't be a perfectly satisfactory solution?.....I mean, why is nobody on the web doing it like this?

The big killer for Phase shift full bridge with high voltage output is the high ringing that you get across the output diodes due to the high leakage inductance...this is a ring with a frequency near the switching frequency, so is difficult to snub out....you don't get this problem with plain full bridge and transformer interleave wound to reduce leakage inductance.

 

[mtwieg]

But supposing that this was done with 4 paralleled CCM boost PFC's.....into 4 paralleled plain full bridge converters...is there any reason why that wouldn't be a perfectly satisfactory solution?.....I mean, why is nobody on the web doing it like this?

First of all, be careful about how you describe the converter. A full bridge can be phase shifted or PWM controlled, and both can be "hard switched." The phase shifted bridge however can usually be made ZVS over a limited line/load range. But they're all basically the same converter, just with slightly different control schemes.

The big killer for Phase shift full bridge with high voltage output is the high ringing that you get across the output diodes due to the high leakage inductance...this is a ring with a frequency near the switching frequency so is difficult to snub out....

It should be nowhere near the switching frequency, unless your parasitics are enormous for some reason. Like factor of five lower at least.

you don't get this problem with plain full bridge and transformer interleave wound to reduce leakage inductance.

A properly operating ZVS PSFB should have much less high frequency ringing than a hard switched FB, since it's hard switching that excites those sharp edges.

 

[treez]

First of all, be careful about how you describe the converter. A full bridge can be phase shifted or PWM controlled, and both can be "hard switched."

thanks, I hear what you say, but do you agree that if the Phase shift full bridge is hard switched then it has much more switching loss than the plain full bridge?

This is because in the PSFB, if it hard switches, we get horrendous effect of the higher fet switching ino the reverse recovering low-side anti-parallel diode, which gives an enormous shoot through current, and this doesn't happen with a hard switched plain full bridge smps, because the plain full bridge never has fets switching on into the reverse recovering diode because it doesn't have the primary circulating current seen in the PSFB?

 

[mtwieg]

thanks, I hear what you say, but do you agree that if the Phase shift full bridge is hard switched then it has much more switching loss than the plain full bridge?

Perhaps., let's see. For both, the waveform applied to the transformer primary should be exactly the same, so the primary current should be exactly the same. The only difference is the switching sequence of the FETs used to achieve the waveforms.

In a PWMFB, a FET commutates with its diagonal partner. Shoot through is not possible. Turn on is not ZVS, so energy is lost due to Coss. On turn-off, the primary will softly reverse polarity due to the magnetizing energy. It will clamp to the rails, and some of that energy will be recovered in the DC bus. But the clamping doesn't last long, and the primary voltage slowly rings down, and all the remaining energy is lost.

In a hard switched PSFB, FETs commutate with their complementary partner so things like shoot through are possible if your dead time is way too short. But you don't get ZVS, so you still get losses from Coss. The primary voltage is always driven to the rails, so the magnetizing and leakage energy is always recovered.

For a soft switched PSFB, it's the same as above except you increase your dead time so that you get ZVS, and thus your Coss losses are eliminated. If dead time is too long, then you end up with behavior much like a hard switched PWMFB. No worse, as far as I can tell.

This is because in the PSFB, if it hard switches, we get horrendous effect of the higher fet switching ino the reverse recovering low-side anti-parallel diode, which gives an enormous shoot through current, and this doesn't happen with a hard switched plain full bridge smps, because the plain full bridge never has fets switching on into the reverse recovering diode because it doesn't have the primary circulating current seen in the PSFB?

If a PSFB is actually hard switched, then the diodes are never given time to conduct, period. Shoot through is possible, but only if you screw up badly.

 

[treez]

Perhaps., let's see. For both, the waveform applied to the transformer primary should be exactly the same, so the primary current should be exactly the same.

Thanks, but I am sure you agree that the current waveform in a PSFB and a Plain Full Bridge are very different, and this is due to the freewheeling current, and the lower di/dt on the transitions with the PSFB

Here is LTspice simulations of PSFB which show it working...

LTspice simulation is called....
Phase Shift Full Bridge converter analysis

- - - Updated - - -

Also, please note the attached simulation, which shows a plain full bridge SMPS compared with a non-ZVS phase shift full bridge SMPS.
The non-ZVS PSFB has switching power dissipation peaks which are 5 times higher than the plain full bridge.....this appears to confirm that the PSFB, when in non-ZVS behaviour, has far higher switching loss than the plain full bridge.

(LT spice simulation attached which confirms this...)

The comparative LTspice simulation is called
"Phase Shift Full Bridge_vs_Full Bridge"

 

[Orson Cart]

@ Treez, the phase shift full bridge can be used for HV outputs, e.g. 400V just need good SiC diodes 1200V and snubbers...

At 35kHz the hard switching full bridge is on a par with the PSFB in terms of total losses....above about 40kHz the PSFB is better (to about 100kHz) above 100kHz a full resonant converter seems to be best (lowest RFI, good efficiency, lowest stress on o/p diodes, no mosfet stress at light loads).

 

[treez]

the phase shift full bridge can be used for HV outputs, e.g. 400V just need good SiC diodes 1200V and snubbers...

SiC diodes have very high forward voltage, and since the current in the diodes is actually more with a PSFB than a plain full bridge, we wouldn't want to use sic diodes.

 

[mtwieg]

Thanks, but I am sure you agree that the current waveform in a PSFB and a Plain Full Bridge are very different, and this is due to the freewheeling current, and the lower di/dt on the transitions with the PSFB

Right, I forgot that the net magnetizing current is identical, but the primary/secondary currents are different, my mistake.

The non-ZVS PSFB has switching power dissipation peaks which are 5 times higher than the plain full bridge.....this appears to confirm that the PSFB, when in non-ZVS behaviour, has far higher switching loss than the plain full bridge.

Note that the switching loss in the PWMFB depends on the voltage at each leg just prior to turn-on. Since the primary ringing in your PWMFB is very underdamped, this means that depending on the exact duty cycle, that voltage may be near either supply rail, and you may actually see ZVS if you're lucky, and your switching losses will be misleadingly low (or high, if the voltage happened to be at the opposite extreme). But it's not plausible to see that ZVS consistently, so you have to take your switching loss measurements with a grain of salt.

Anyways, with regards to evaluating the simulated switching losses, just looking for the peak power is not the way to go about it. You should sum the power dissipation of all four FETs in each circuit then look at the energy of those spikes, not the peaks. When doing that, I find I get 19uJ for the PSFB (with no ZVS) and 13uJ for the PWMFB. So yes, if your PSFB is very poorly implemented, then it can have higher switching losses. If I adjust it to give ZVS on the FPSFB, I get less than 1uJ.

If should be noted that in both of your example circuits, the difference in their total dissipation in the FETs is mainly due to increased conduction losses in the PSFB (those extra circulating primary/secondary currents), not switching losses. Using much lower Rdson FETs would likely make PSFB come out ahead, especially with proper tuning.