Bipolar FET drive in LLC converter needs many components?

[Faringdon]

Hi,
Its just dawned on me , the sheer amount of components that are needed to achieve Bipolar FET gate drive (+15V, -5V)in an offline LLC converter. The attached shows the minimum it can be. (jpeg and LTspice attached)
An isolated gate drive IC (such as eg 2ED2106S06F) could be used  so that the pulse transformer (and the 'upstream' gate driver) could  be omitted, but these isolated drive ICs can potentially suffer from malfunction due to negative voltage spikes on the switching node (the ICs “hot” side ground).

And of course, whatever componentry you use for the hi side FET...you must use exactly the same for the low side FET, so that the propagation delay is the same (preserving dead time).

Can you think of a way to reduce component count here? 

[T3sl4co1l]

Everything left of the gate drivers can be replaced with a digital isolator e.g. ADuMxxxx.  Indeed I did this ca. 2011 for induction heater power supplies running up to 400kHz, 5kW per (full bridge) board.

Even more can be replaced with a more tightly integrated driver (isolation + drive + fault detect) though this might be a cost or availability tradeoff.

Tim

[Faringdon]

Indeed I did this ca. 2011 for induction heater power supplies running up to 400kHz, 5kW per (full bridge) board.

Thanks, may i ask what was the input voltage for this?

Everything left of the gate drivers can be replaced with a digital isolator e.g. ADuMxxxx.

Thanks, the cheapest ADuMxxxx that exists with >=100V/ns CMTI is ADuM4120 at $5.5

Would you say that the 1EDi20N12AF, at $2, would be just as good?, (since we dont need the 5000V isolation)

ADum4120 datasheet
https://www.analog.com/media/en/technical-documentation/data-sheets/adum4120-4120-1.pdf

1EDi20N12AF datasheet
https://www.infineon.com/dgdl/Infineon-1EDI20N12AF-DS-v02_00-EN.pdf?fileId=5546d4614755559a014790299add6112

(BTW, for some reason, the ADuM4120 datasheet doesnt give min and max switching frequencies)

[T3sl4co1l]

Indeed I did this ca. 2011 for induction heater power supplies running up to 400kHz, 5kW per (full bridge) board.

Thanks, may i ask what was the input voltage for this?

Up to 650VDC

Would you say that the 1EDi20N12AF, at $2, would be just as good?, (since we dont need the 5000V isolation)

Yes, and integrates driver besides.

(BTW, for some reason, the ADuM4120 datasheet doesnt give min and max switching frequencies)

Min pulse width 50ns

Tim

[Faringdon]

Thanks, and the minimum frequency is probably 1Hz or so?

[wraper]

Thanks, and the minimum frequency is probably 1Hz or so?

What minimum frequency? There is none. Not to say frequency is irrelevant in this case.

[Faringdon]

Thanks,
Actually , what do you make of the attached LLC FET drive, which maintains the full drive voltage at the FET.? (so it can drive SiC at Vgs = 18V)

No Hi side supply, just a gate drive transformer.
Can you see any gremlins with this?
(LTspice and jpeg attached)

[T3sl4co1l]

Still sucks.

The one advantage for GDT is only needing one of them, with dual windings.

Most SiC want reverse bias, making this unsuitable anyway.  There are a range of types available, of course.

GDTs aren't great above a few kW, at least with Si; I suppose that range might be extended given the lower Ciss of SiC, but then the precision of Vgs required, and negative bias if applicable, makes that a bit dubious again.

Tim

[Faringdon]

Hi,

Most SiC want reverse bias,

Thanks, i take it that   you are referring to Bipolar gate drive here(?)....
The attached  (LTspice and jpeg) has bipolar FET drive…but when you trace out the gate drive loop, you are going through more components…….through each of the split caps….and  it looks like there will be significant  gate drive path inductance…leading to the problems that bipolar drive was put there for in the first place. I think if the gate drive is very well damped, then bipolar drive looks OK….but not otherwise. I can see bipolar drive being good with IGBTs for drives, which have low switching frequency, and so it can be well damped.
The  PNP turn off circuit, when put right next to the FET in question, does appear to be the best, low impedance way to guarantee the FET staying off. And of course, you can’t have “PNP turn off”  AND “Bipolar FET drive” together.
Would  you agree?
I think bipolar drive may be OK though, if you could have the bipolar driver and components done in SMD and right next the fet being driven, with very tight layout…..but on a big power board, you cant have SMD on the board, as you know, the SMD has to be on a daughter board…..so you’ve got to route off to that, and then you get more stray  layout inductance.
I think Bipolar drive is greta in principle….but on a big power board, when you come to do it…..it doesn’t look better than a PNP turn off circuit.
I think the emergence of “Active miller clamp” on the modern gate drive IC’s is testimony to the fact that Bipolar gate drive has problems, and other ways need to be looked into……the  “Active miller clamp” method , of course, is basically a PNP turn-off circuit.
Would you agree?

[T3sl4co1l]

Yes, bipolar, whatever you like to call it.

No, I wouldn't agree.  Nothing is specific to using a PNP, that's just one among many ways to achieve low drive impedance in the low state.  Maybe Miller clamp is adequate, maybe reverse bias is needed, maybe both.  I don't know what these devices need.

Tim

[Faringdon]

Thanks, i thought Bipolar FET gate drive was good, until i schem'd it...and looked at the gate drive loops....(for FET  ON and OFF)
The problem with bipolar FET gate drive is that the hi-side fet driver chip ground is not the same ground as the hi-side FET source node(!!!)

The hi-side  FET source node (= switching node) is the "split capacitor" point, of the hi-side bipolar supply rails............so when you turn on your hi-side  fet.....you are doing so from the hi-side fet driver chip...which is referenced to hi-side fet driver chip ground...which is not  "hi-side fet ground"........so your turn on comes , as always, initially, from the fet driver Vcc cap...(which is not referenced to the fet source)........so your ON drive current comes out of the vcc cap..........then goes to the driver txtor in the driver chip.....then out to the fet  gate.....then out thru the fet source......then goes along the  "split cap node" trace......where it finds the lower cap of the "split cap node"...it then goes thru that lower cap.....and then along the fet driver ground trace...until it gets back to the vcc cap....

......now try routing that all nice and tight on a 500W+ SMPS power board. (as you know, no SMD on a high power main board, and if daughter boarding...youve got all the trace inductances to get on and off that daughter board)

....ditto the grief for the turn off  gate drive path.

..Thats why i dont like bipolar FET gate  drive for high power SMPS.....just prefer a  small "turn off PNP"  circuit  right at the fet gate....nice and tight.

For a well damped, low frequency electric drive circuit however, and bipolar rails can work well...plenty of damping in the gate drive path.

[T3sl4co1l]

Back when I did some industrial IGBT modules, I used a, IXDD614CI I think it was, at 30V (+/-15) with a zener to set the emitter voltage and extra caps to handle the gate charge (100s nC for these big beasts).  Think it was 2 x 68uF aluminum polymer plus a couple 1uF ceramic chips, bypassing +15/0/-15 together.  Risetime was dominated by the interconnect anyway (at least stray inductance of the connectors/pins and internal connections in the module, but also connecting leads when boards were not placed directly over modules but wired in with extra twisted pair(!)), besides which these modules weren't exactly speed demons and topped out at 50kHz or so (at reduced ratings) so it was all rather overkill, really.

That was something like 150kW for a complete inverter module, using two H-bridge modules in parallel.  (Same company, different product from the 5kW MOSFET module.)

Length of the drive return path doesn't mean anything, you can just use more ceramic caps in parallel to make a wider connection.  It's the ratio of length to width that matters.  Not like you're getting it much below 10nH anyways (if these are THT MOSFETs).

I also tried running the 5kW MOSFET module (+12/0V gate drive) with IGBTs, which almost worked at 20kHz, but alas, almost.  The ensuing arc flash removed a few pesky traces from the PCB in the vicinity of the blown transistors.  We did a custom 600Hz unit though, which used IGBTs just fine at something like 10 or 20kW, using otherwise the same [5kW PCB].

Tim

[Faringdon]

I also tried running the 5kW MOSFET module (+12/0V gate drive) with IGBTs, which almost worked at 20kHz, but alas, almost.

Thanks, thats interesting.....it would also be interesting to hear if your 12V/0V drive included PNP turn off right by the txtor?
Also, i appreciate that IGBTs need a negative turn off due to their "tail current", which as you  know, MOSFETs fortunately dont have.

This App Note
https://www.semikron.com/dl/service-support/downloads/download/semikron-application-note-unipolar-driver-voltage-en-2015-06-03-rev-02/
on page 1,  tells of how IGBT switching delay parameters change between unipolar/bipolar drive.
This isnt the case with FETs, as you know.

but wired in with extra twisted pair(!))

...I must admit, this sounds like a really good idea, on a big power board, to get a gate drive signal from a daughter board to a TO247 FET.

[T3sl4co1l]

No PNP.  Why bother?

Was FAN3121TMX I think.

Twisted pair isn't a good idea for MOSFETs.  Check the gate circuit RLC equivalent.  You'll see why.

The big IGBT modules look something like 0.2uF + 2R so can tolerate quite a bit of inductance in the loop.  MOSFETs not so much.

Tim