Why is our Bootstrap high side drive circuit blowing up?

[treez]

Hello,
Why is the damping resistor in our Bootstrap high side FET drive blowing up?
Schematic attached.

The boss says that when the bootstrap diode of this circuit was using a silicon carbide bootstrap diode, the circuit worked OK. Then that SiC diode went obselete, so he changed to the US1M ultra fast bootstrap diode, and ever since then, the damping resistor keeps blowing up. (the bootstrap diode footprint is SMB, and we are limited to using whatever fits on that footprint for the replacement)

The SMPS being driven is a 5kW Phase Shift Full Bridge converter with an input voltage of 564Vpk (the output of a three phase rectifier)

US1M diode datasheet:
**broken link removed**

ADuM4223 datasheet:
https://www.analog.com/media/en/technical-documentation/data-sheets/ADuM3223_4223.pdf

UCC27322 datasheet:
https://www.ti.com/lit/ds/symlink/ucc37321.pdf

{do you reckon its got something to do with the bootstrap capacitor being too large (10uF), and so at start-up, when this cap isn’t yet fully charged up, the bootstrap diode is still significantly conducting when it suddenly gets reverse biased, and then the reverse recovery damages the said resistor?)

 

[FvM]

Have been using US1M in similar circuits with 1R mini-Melf series resistor, Udc = 650V. Mini-Melf Pulse rating is slightly higher than 1206 but almost comparable.

A possible circuit problem could be large switching node undershoot due to bad layout.

 

[treez]

thanks

Have been using US1M in similar circuits with 1R mini-Melf series resistor, Udc = 650V. Mini-Melf Pulse rating is slightly higher than 1206 but almost comparable.

Do you think for a few nanoseconds, the damping resistor actually sees the full 530 Volts in the "reverse recovery" time of the diode?

Sorry I cannot go and measure it as the above is the only info the boss has given me, I don't have access to the circuit. (he's actually just asked me to make him a diode reverse recovery circuit so he can compare the reverse recovery transient of several alternative diodes , for use in place of the US1M)

I also noticed that the ADuM4223 isolator has a transient immunity of only 25V/ns, and the switching node of the Phase Shift Full Bridge SMPS could well see its switching node transition at >25V/ns?

Surely , I reckon the Infineon 1EDI20N isolated gate driver is the only isolated driver that should ever get used with SMPS..you agree?…….its transient immunity of 100V/ns is the best in class.
1EDI20N isolated gate driver:
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d4614755559a014790299add6112

 

[joao.caracas]

treez,
- Have you tested reducing R1 to 1R?
- What is the voltage rating and type of C2?
- What is the gate charge of MOSFET M1?
- What is the pwm frequency?

 

[FvM]

Isolator transient rating is irrelevant for the bootstrap resistor at first sight.

I don't know how much transient voltage does the series resistor see. I can just confirm that in my design, we had no noticeable heating of the series resistor (according to IR camera). If you try to keep pulse ratings of regular thin or thick film resistors strictly, you walk on rather thin ice nevertheless. That's why I put in a better specified mini-Melf, just in case.

Presumed diode Qrr is mostly independent of resistor value, a lower resistance would in fact reduce pulse power.

By the way, what's your sequence for initial bootstrap precharge? It might play a role if the resistor already fails during start-up.

 

[treez]

By the way, regarding the case of potentially damaging negative transients making the ADuM4223's output_side GND pin going well below the input_side GND pin, (due to ringing related to the fast transitioning switching node) do you think the attached R/Diode protection circuit (ringed) is a good way to solve this?
-This protection is an idea gotten from page 16 of the attached NCP5106 bootstrap driver IC datasheet.

By the way, what's your sequence for initial bootstrap precharge? It might play a role if the resistor already fails during start-up.

Sorry i dont know, the above is the only info the Boss lets me see.-And He only told me this to put into context the diode reverse recovery test circuit that he's asked me to build.

I don't know how much transient voltage does the series resistor see. I can just confirm that in my design, we had no noticeable heating of the series resistor (according to IR camera).

...thanks for this observation, i will take it from this that due to the fact that this resistor didnt get hot that it is actually overvoltage that was causing the resistor to fail.

- - - Updated - - -

................................ooo.............................................................

treez,
- Have you tested reducing R1 to 1R?
- What is the voltage rating and type of C2?
- What is the gate charge of MOSFET M1?
- What is the pwm frequency?

...............sorry i dont have acces to circuit
C2 is 10uF, 1206, x7r, 25V.
M1...sorry i dont even know what part this is.
PWM frequency, sorry dont know.
..As discussed above, i was only told what the problem was to put into context the reason they are asking me to build a diode reverse recovery test jig.
...................................000............................................................

 

[FvM]

The protection circuit is intended for level shifting drivers like IR2110 or NCP5106 and useless or even problematic for the present isolated gate driver design. It won't protect the bootstrap supply circuit at all.

 

[treez]

Thanks,
Do you think its worth putting in the (ringed) charge-up resistor (schematic attached) , which can pre-charge the bootstrap capacitor before the bridge starts switching?….this means that the bootstrap cap will be fully charged when switching starts, and so the current in the bootstrap diode when it first gets reverse biased is likely to have stopped, -meaning that there will be no reverse recovery transient in the bootstrap diode.

As you know, in steady state operation, the bootstrap capacitor is highly charged, and the bootstrap diode only conducts into the bootstrap cap for 100ns or so…meaning that it has stopped conducting current when it gets reverse biased, meaning no bad reverse recovery transient for it.

- - - Updated - - -

Isolator transient rating is irrelevant for the bootstrap resistor at first sight.

At first sight, no, but do you agree..if an overly high dv/dt makes the ADuM4223 switch when it shouldn't, then the bootstrap diode may get "reverse dv/dt'd" when it is conducting near its maximum current, which could cause a bad reverse recovery transient...you agree?

 

[FvM]

The "charge-up resistor" will only supply µA current and can't win against driver quiescent current. Don't believe that it produces more than a few volts.

Parasitic switching due high dV/dt can be a problem, but involves a risk of output transistor damage in the first place.

 

[Vbase]

The diodes at the gates of the power mosfets are switching the mosfet on fast and off slowly. I find it very unusual and can't think of a case that it is beneficial.

 

[treez]

thanks, sorry I put those the wrong way round in the schem, not the board. (please excuse not reloading as it isnt central to the question)

 

[Vbase]

If the damage to the resistor is a tiny black spot and the resistor is OC then it is from high voltage. If the resistor is OC and no visible damage then it's likely fused from high current.
My guess is that the resistor damaged from high current and it needs to be increased to 10R.

 

[treez]

thanks, or keep it 4R7 but just make it say two in series or a melf like FvM said?
You agree that turning the bottom fet on at first at startup to charge up the boot capacitor is a good idea.....so that when the high fet eventually gets turned on the boot diode is not conducting and so doesnt have a reverse recovery problem?

 

[FvM]

I don't think that the resistor is damaged with the diode is in forward bias. I don't see the problem even about 250 µF total bootstrap capacitance.

The reverse recovery explanation sounds more likely to me. Respectively increasing the series resistor does no necessarily reduce the pulse load, as mentioned before.

 

[treez]

what I mean is that the bootstrap diode should actually never have a reverse recovery in steady state operation of the SMPS...because it wont be conducting into the bootstrap capacitor at the point that its dv/dt suddenly increases with the onset of reverse bias. If a diode is hit with reverse bias when it isn't conducting, then it sees no reverse recovery...you agree?

- - - Updated - - -

Thankyou all, I believe that these kindly donated posts allow us to make a “law of bootstrap hi-side gate drivers for mains voltages and above”

“The law is that you must use a SiC bootstrap diode, even if that means having to use a TO220 SiC package so as to get a low Vf.”

Do you agree with this law?

Clearly reverse recovery of silicon ultra fast diodes renders them useless for this purpose, since the reverse recovery can blow up the damping resistor, and if one uses too small value of damping resistor to mitigate the problem, then the reverse recovery unwontedly discharges the bootstrap capacitor too much.

 

[FvM]

Nothing has been said about duty cycle yet. Reverse recovery can be significant with short low side on-time.

 

[treez]

yes I agree, but do you now agree that the law of post #15 applies....

“The law (for mains voltages and above) is that you must use a SiC bootstrap diode, even if that means having to use a TO220 SiC package so as to get a low Vf.”

Do you agree with this law?

Clearly reverse recovery of silicon ultra fast diodes renders them useless for this purpose, since the reverse recovery can blow up the damping resistor, and if one uses too small value of damping resistor to mitigate the problem, then the reverse recovery unwontedly discharges the bootstrap capacitor too much.

One could use a melf resistor , but who wants to have a high pulse current being discharged out of their bootstrap capacitor?

 

[dick_freebird]

You have about three ways to fail that diode. Voltage,
current and power (energy, pulsed or cumulative, as
heat).

Nuances of design may matter here. SiC diodes can
hang tough, leakage- and storage-time-wise out to
past 300C. Silicon may not, comes down to cases.
Pulsed power levels in the switching events may spike
internal temperatures well above baseplate. If this
happens the diode may have positive temperature
feedback. A silicon die may be smaller than a SiC die
for same current rating, as conductivity modulation
may be stronger - but the price paid for this is usually
storage time, and storage time beyond rated junction
temp is going to be a mystery, its trajectory / limits.

Deprocessing the packaged device and inspecting
is going to be the most informative thing you can do.
The rest is confined to speculation unless someone
has been down this exact path ahead of you, and is
here to share.

It's not too hard to find SiC and GaN diodes now; is
your focus misplaced or obsolete? Why no more effort
to find similar / superior SiC diodes elsewhere? Or an
"over-rated" silicon type with more power handling, as
a diagnostic substitution, trying to determine the mode?

 

[treez]

Thanks, its not the diode that's failing its the damping resistor.

Genesicsemi.com make sic diodes in SMB case. But they have high Vf....need to use to220 sic diode to get the vf down.

 

[dick_freebird]

Maybe a larger body resistor, or a thick film instead of thin
film, might be better able to take the peak pulse power ?