Phase Shift Full Bridge SMPS is massively over-hyped?

[treez]

Hello,
The Phase Shift Full Bridge (PSFB) converter is a hoax. –The LTspice simulation attached of a “plain” Full Bridge SMPS bears this out.
PSFB claims to be able to reduce switching losses compared to the “plain” Full Bridge SMPS.

However the PSFB does have Turn-OFF switching losses, just as does the “plain” Full Bridge SMPS…so “no cigar” there for the PSFB.

The PSFB does have zero turn-ON switching losses, -but the “plain” Full Bridge SMPS has very minimal switching losses anyway. The only turn-ON switching loss of a “plain” Full Bridge SMPS is that associated with the 1/2CV^2 loss due to the discharge of the FET’s Drain-source capacitance at turn-ON. –But this is a small switching loss.
The “plain” Full Bridge SMPS has no “overlap” of FET voltage and current at turn-ON because the leakage inductance of the Full Bridge transformer prevents the current from rising up quickly enough to do that.

The downer for the PSFB is the large amount of circulating current in the primary of the PSFB transformer, -this causes more conduction losses than the “plain” Full Bridge SMPS…so definitely “no cigar” for the PSFB here.

The PSFB is simply a waste of time, its advantages are too insignificant to make it worth the extra expense. PSFB’s don’t even manage to get zero turn-ON switching losses when lightly loaded. Waste of money…do you agree?

PSFB actually has more dissipation loss in the secondary side diode snubbers than a “plain” Full Bridge SMPS, -Due to the enhanced leakage inductance of the PSFB tranformer. Why is anyone using PSFB?
You can’t even use PSFB at high frequency (like most resonant converters), because the leakage inductance robs you of duty cycle at higher switching frequencies.

 

[Easy peasy]

The PS FB is used in the 100's of thousands out there for electric forklift chargers running off 3 ph mains and typically 6kW of output, at 100kHz, the turn off losses are dictated by the speed of the gate drive, the turn on losses are near zero for 20% - 100% load, the EMC is much lower due to the reduced dv/dt on the o/p diodes (due to the series L on the pri side to extend lossless switching behaviour to 20% load), a standard hard switched bridge could never compete purely due to turn on losses and EMC and o/p diode stresses....

 

[treez]

As you know, there are no "overlap" (Vds & Ids occurring together) turn-on losses in a full bridge SMPS. -Just the discharge of the Cds through the fet, but this is a small loss. The diode turn on in a full bridge is also at a reduced dv/dt due to the leakage inductance of the full bridge transformer, -admittedly it is not as slow a dv/dt as the PSFB , but it is slow enough to be satisfactory, and there are many full bridge smps's out there working at 10kw plus to prove this.

The "extra leakage inductance" added with the PSFB means that diode snubber losses with the PSFB are higher than with the Full bridge.

Also, the far higher RMS primary current with the PSFB makes it only very marginally more efficient than a plain full bridge smps.

to get zero turn on losses with the PSFB, the added series inductance is unreasonably high , resulting in extra losses in the secondary diode snubbers........also, since the extra leakage makes the ringing at a frequency which is not far enough away from the switching frequency, it is difficult to snub this lower frequency ringing. This is less of a problem with the plain full bridge.

Another thing we speak of is the duty cycle reduction which happens with PSFB's with significant added extra leakage inductance.....so getting the turn on losses down to 20% load is not without a heavy price.

I know that there are 100's of thouands of PSFB's out there running at several kw's...but each of those will have a hard switching PFC stage ahead of it....yes, as you know, a hard switching boost converter(s), with lots of hard switching loss, and lots of diode reverse recovery loss...nobody ever says that that is a show-stopping problem. Many people speak of hard switched full bridge smps's as if they are the end of the world, then they go and do a resonant converter with a hard switching PFC upstream of it...........the consultancies who "nonsense talk" their customers into agreeing to a resonant converter say nothing of the hard switched PFC stage(s) that they will be putting in with it.

 

[Easy peasy]

actually there is some overlap in a std full bridge, and the snubber energy at turn on to be absorbed, compared to the extra caps across the fets in a FBPS, try running a hard switched FB at 100kHz...!!! at 6kW, at 650V bus, if you ever do you will see the losses and the RFI generated by the hard switch full bridge..!
For 3 phase 6kW there is no PFC front end as no standards require it...!

 

[treez]

For 3 phase 6kW there is no PFC front end as no standards require it...!

As you know, when electric vehicles get very commonplace, there will be a change in the regulations, and for that market, there will be full 0.99 pfc required for all the PSU's (chargers) that run off single or three phase, no matter what the power is in kw's

actually there is some overlap in a std full bridge

..at turn on its hardly any, run the simulation of post#1 and see for yourself, the leakage inductor cuts most of it out.

As you know, an active boost converter PFC stage cannot be placed downstream of a three phase rectifier, to get PFC with three phase, you have to have each phase running through a full bridge and use the pfc(s) there.
As you rightly say, in your case you do not need pfc, but electric vehicles do now. So we would not see the 650V bus that you speak of, all that our full bridges will see is the 390V output of the PFC stage.

if you ever do you will see the losses and the RFI generated by the hard switch full bridge..!

As you know, we don't run off 650v in commercial domestic vehicle market, and we do see all the RFI that you speak of due to our very hard switching pfc stages, and its nothing of great concern. We had a consultancy approach us and tell us we should pay them 100's of thousands for a special resonant converter that only they could do properly.....they told us that hard switching was not possible at 7kw...then we asked them how does the pfc stage work then, and they went quiet and left our offices soon afterwards.

 

[Easy peasy]

I have never seen a hard switched FB at 6kW (or anywhere near that power) running off rectified 3 mains any where near 100kHz, 30kHz at best... boost converters are far different as they process only a fraction of the thru power...
there a re plenty of 3 phase boosters out there with no neutral (i.e.not 3 x single phase adaptations) but they are nt used even now for a lot of EV chargers as six diodes wins out for size and cost and the pf > 0.9

 

[treez]

I have never seen a hard switched FB at 6kW (or anywhere near that power) running off rectified 3 mains any where near 100kHz,

I understand, though as you know we aren't running full bridges off the output of a three phase rectifier.

boost converters are far different as they process only a fraction of the thru power...

Sorry but surely you agree that the boost converter PFC is one of the most hard switching topologies out there.....during the switching transistion there is rapid discharge of the Cds capacitor, also horrendous reverse recovery, as well as overlap of v and I during the switching transition.

I don't know what you mean by saying it only "Processes" a fraction of the thru power.....during the ON cycle, and the OFF cycle, the current is passing through semiconductor.....either the fet or the diode (as well as going through the power inductor all the time if in CCM)......as such , the boost converter is processing power 100% of the time.

The hard switching reverse recovery of the boost pfc diode is one of the most highly EMI unfriendly acts in the whole of the SMPS world.

 

[Easy peasy]

If you work out the actual power processed by a booster it is only a fraction of the total input power even tho the fet turns on into peak mains current. A std hard switched booster is unreliable at 3kW for the reasons you mention, most are split or use softer diodes (or SiC) or have a lossless snubber to control Irev thru the diode and reduce sw losses in the main switch.

 

[treez]

so the following 3kw PFC , offered up by Infineon, is reliable because it uses a sic diode...
**broken link removed**


I don't know what you mean by "process" power, as you know in a boost converter CCM PFC, the current into the boost converter always flows in silicon, whether it be the fet or the diode.

 

[Easy peasy]

yup, only because of SiC diode, OK determining the power processed by a booster is non-obvious, but to explain it simply, when the fet is off the current is pushed thru the boost diode by the volts from the choke and the volts on the mains at that instant, so at the peak of the mains the boost choke is doing very little work, as it has to add only a few volts, at the zero xings it is adding lots of volts but very little current, it is possible to analyse and calc the exact amount of power processed by a booster, it is about 30% of the total input power, which is why a booster is efficient, the fet does not carry all the input current all the time (as an isolating converter would have to) and the diode carries only the average o/p current, it is a function of how much AC the converter has to make to do its job, an isolating converter has to make 100% AC to cprocess its power, a buck converter at 100% on has to make 0% ac, as does a booster at 0% on (all in diode), so you see for a booster the lower the average on time the less power is processed (AC made)...

- - - Updated - - -

this also explains why boosters are more efficient at high mains, lower ave on time = less power processed, ignoring the extra i^2 R at low mains...which also contributes

 

[treez]

thanks, I see what you mean. Its kind of a side point though to the theme of the discussion, in spite of being technically accurate. In say a full bridge smps, as you know, only two diagonal fets conduct at any one time, the other two have a 'rest', so we can say that the individual fets of a full bridge don't carry all the input current all the time....they share it between them. I think when we speak of resonant converters, we accept the they only reduce switching loss and emi, and the boost converters "less then 100% power processing" feature doesn't really reduce its switching losses. The boost FET suffers switching loss's every cycle.
Also, in a full bridge smps, each individual diode only carries half of the average load current, as you know. (as in the top post's simulation)

I know you may say that a boost converter on very low duty cycle has the fet "processing" very little of the power, but I believe that's not really relevant to the subject that boost converter pfc stages, being very hard switching, are also very commonly used at high powers, up to 3kw as you say also.

And this harps back to (not yourself), but the raft of consultants telling us that "you cant do hard switching stages above 1kw" etc etc...hang on, the hard switching boost converter PFC does it regularly. Just because it doesn't "process" all the power, doesn't make it any less hard switching. I think we are going off-topic on that one, I am not saying it is inaccurate by any means.

 

[Easy peasy]

The underlying message is: Its easier to make a 3kW booster than a 3kW isolated converter because a booster is structurally simpler and you don't need components to process all of the 3kW (for e.g.) that is required, as you do for a full converter.

 

[treez]

Agreed, and as you know, a boost converter at 100khz has just as much switching loss as a full bridge at 100khz (I am referring to an individual transistor in each one, I appreciate the FB has more transistors).
Also, a PSFB at 100KHz would have just the same switching loss as a full bridge at 50khz (but please see below), because the PSFB has full switching loss on the turn off transient.
The difference between them, is the turn on switching loss, but the full bridge has mild switching loss there, caused in the main by just the discharge of the energy in the Cds capacitance.
So in fact, a 100Khz PSFB would have the same switching loss, in fact, as a Full bridge working at around 75-80khz.

 

[Easy peasy]

I see your line of argument, but actually the turn on losses at 100kHz at 400 - 600V DC bus are higher than you might expect for a std H bridge, there is a contribution from the recovering o/p diodes (and their snubbers) that you can't ignore. Actual practical implementation shows there is <1.5W turn on loss per device in a PSHB with zero volt turn on, compared to 10W or so for a std H bridge, this is largely because you cannot turn the fets on really fast and hard in a std H bridge as you will hit the o/p diodes really hard in doing so, upping losses and RFI, + the self capacitance discharge. Upping the leakage above a useful minimum value does not help as it hits the o/p diodes harder requiring snubbing (even for SiC). It is easily 10W per device turn on losses or higher trying to switch a std H bridge at 100kHz in a conventional rectifier arrangement, which is why it is very rare to see.
By way of example, reverse recovery and snubber currents causing a 5A ave peak for 100nS in a fet on a 400V bus, @ 100kHz, gives 10 watts per device on the mosfets at full power, assuming they both turn on perfectly at the same time and share the 400V, else one will get hotter than the other.

 

[dick_freebird]

To me the value in a phase shift bridge is that you can
get to zero, or either side of it, while a fixed phase bridge
runs into minimum on- and off-time control problems at
either end of duty cycle. So more suitable for extremely
high boost ratios than a fixed phase bridge perhaps, in a
boost converter where duty cycle goes directly to boost
ratio.

If you aren't outside maybe 0.25:1 - 4:1 Vo/Vi range then
I doubt phase shifted approaches bring much to the party,
you can get the control range you need with regular parts.

 

[treez]

So more suitable for extremely
high boost ratios than a fixed phase bridge perhaps

thanks but sorry I though it was the other way round, PSFB isn't good for high ratios because a PSFB suffers 'duty cycle loss' due to its (designed) higher leakage term in the transformer.?

- - - Updated - - -

I see your line of argument, but actually the turn on losses at 100kHz at 400 - 600V DC bus are higher than you might expect for a std H bridge, there is a contribution from the recovering o/p diodes (and their snubbers) that you can't ignore

the simulation in the top post shows that the normal leakage inductance in the full bridge transformer prevents the reverse recovery current (of the secondary diodes) from flowing in the full bridge FETs...I deliberately implemented ridiculously bad diodes in that simulation to show this point.
------------------------------------------------
I know we have said that the boost converter doesn’t “process” the power as much as other hard switched converters, however, comparing the boost converter and the full bridge converter (as in attached simulation), the semiconductor with the highest power density in it is the boost converter diode, so as you know, we cannot write off the boost converter PFC stage as some kind of power dodger. (not that we were but lest we forget)
The LTspice simulation attached compares the boost converter and full bridge converter doing a 200vin to 400vout conversion at 400watts (both at 100khz).

Back to PFC's, the Boost converter PFC’s diode gets hit very hard!….with both conduction loss, and reverse recovery loss if not sic..if sic then extra conduction loss due to sic’s high forward voltage.

In the simulation the boost inductor carries twice as much current as the output inductor of the full bridge. The boost converter PFC process’s significant power I believe. It does this as a hard switching converter, and does it up to several kw’s. This proves that hard switching smps’s can do several kw’s.

We often hear consultancies telling us we must pay them highly for their super duper resonant converter because "nothing else will do", and certainly not hard-switched. I think the PFC boost debunks this argument. Also, resonant converters need a lot more maintenance if eg a fet goes obsolete then you've often gotta have people around who know how to re-set things like dead times for the resonant behaviour..no doubt the consultancies offer to come and do this for gazillions of fees, often times a good hard switched converter is perfectly satisfactory. At one UV LED place we had hard switching buck led drivers (500w by way of three bucks on a pcb) which were 96% efficient. Admittedly from 48vin.

because you cannot turn the fets on really fast and hard in a std H bridge as you will hit the o/p diodes really hard in doing so, upping losses and RFI

..turning the diodes on quickly causes power losses?

The PSFB has worse output diode snubber losses than a full bridge....especially so since the leakage term is usually higher in a PSFB, making the leakage ring across the output diodes at a lower frequency and closer to the switching frequency making it harder to snub out without incurring more losses.

 

[mtwieg]

..turning the diodes on quickly causes power losses?

Yes, look at a datasheet for a power rectifier and you will see that its reverse recovery charge depends on the di/dt. At higher di/dt, Qrr becomes much larger, and therefore so will the dissipation. The extra leakage in the PSFB causes slower recovery with less losses. The ringing is no more severe, since the leakage inductance is not the cause of ringing in a full bridge (the parasitic inductances in series with your switching devices are).

 

[treez]

the leakage inductance is not the cause of ringing in a full bridge

if you run the simulation in the top post with an ideal transformer, (no leakage), you will see that you get no ringing across the secondary diodes, because the ringing across the secondary diodes was caused by the leakage inductance in the transformer.


PSFB ringing across secondary diodes needs snubbing, that we agree on...and psfb ring across secondary diodes is a lower frequency ring than in a plain full bridge (generally), so it takes more dissipation to snub it, as its nearer the switching frequency and we would all prefer to snub ringing that's well higher than f(sw).

I am sure you agree that a plain full bridge has leakage too....which to an albeit lesser extent, has the benefits that you speak of with regard to di/dt of diode and lessening reverse recovery. The PSFB does not have a monopoly on that advantage...our friend the plain full bridge shares in it too to an extent.....talking to some of the consultancies that we talk to, would make many think otherwise.

In a PSFB, the extra leakage term can cause "duty cycle loss". Also, let us not forgot the very significant increase in primary circulating current that we get with the PSFB... that we all agree causes increased conduction losses compared to full bridge.

 

[mtwieg]

if you run the simulation in the top post with an ideal transformer, (no leakage), you will see that you get no ringing across the secondary diodes, because the ringing across the secondary diodes was caused by the leakage inductance in the transformer.

Oh I thought you were talking about snubbing of the bridge, my mistake.

PSFB ringing across secondary diodes needs snubbing, that we agree on...and psfb ring across secondary diodes is a lower frequency ring than in a plain full bridge (generally)

Sure.

so it takes more dissipation to snub it

I don't agree. This would be in some topologies where the leakage energy is not recovered, but in a PSFB it is not the case. Just to see, I ran a couple simulations on one of your PSFB models while changing the leakage, and the loss in the snubber resistors does not change substantially (assuming the snubber values have been re-optimized for the new leakage values).

 

[Easy peasy]

the simulation in the top post shows that the normal leakage inductance in the full bridge transformer prevents the reverse recovery current

Hello Treez, respectfully, you put too much faith in simulation and then reason from incorrect results, had you build a 3kW converter you would see the high levels of RFI resulting from diode turn off even at modest di/dt (and dv/dt) resulting from modest turn pn speeds of the fets, wind up the turn on of the fets and everything gets much worse, affecting control.

Yes a booster does it, but the diodes are well chosen and they do dissipate a lot of heat at 3kW due to reverse recovery, wind up the turn on on the boost fet and the diode will die unless it is incredibly well heat-sunk, building one would show you this. Hence the turn on resistors in nearly all boosters. Also the effect is isolated to just the boost parts, if you tried to do this through a Tx the RFI would be un-managable and you would never pass EMC, s you have four mosfets and at least two output diodes trying to handle very high di/dt & dv/dt. The duty cycle loss is pretty minimal, about 10% worst case, just change the Tx turns ratio to compensate.
We know that changing the PSFB on the 6kW to a hard switched FB would mean that the unit would overheat and cause all manner of RFi head-aches, that would be practically unsolve-able...

As to the extra leakage on a PSFB, usually there is an extra inductor in series with the Tx, and there are diodes to the rails where this choke connects to the Tx , this mitigates the huge snubbing you think is required on the o/p diodes, all the 6kW converters we see have pretty modest snubbers on the diodes (65V 130A out).