Power Inverter Conversation

Thread Starter

PapaFet

Joined May 15, 2022
6
Hello! I am looking to chat about a power inverter I've designed and built. I say designed but it's really more like adapted working designs for my own purposes. I need some guidance on the failures I'm running into and sites like Electronics Stack Exchange don't seem geared towards a back and forth conversation. I'm not a power electronics pro, just an electrician with electronics knowledge. I also don't have the test equipment necessary to really zero in on the problem(s). I'm hoping that with my descriptions and details that we can talk our way through it.

The Inverter: The design is a "low frequency" pure sine wave type. An Arduino Nano outputs an SPWM signal that feeds a pair of IRF21844 half bridge drivers. Those then control low and high side TLP351 opto gate drivers. The power stage is an N-channel FET H-bridge. I've used both IRFB7545 FETs and IRF3205. Exploding both in spectacular fashion. The output of the FETs goes into a 100 uH choke and then a 3 kVA toroidal transformer. Input is 24 VDC, output is 120/240 VAC 60 Hz. I prototyped it all on perfboard originally to good effect. I wanted it to be modular so I had some PCBs made where each FET is on its own board with an individual TLP351. A 4.7 ohm gate resistor is in parallel with a reverse biased Schottky diode and a 22 kohm gate-source resistor as a pull down. Each high side quadrant has an isolated 12 V power supply for the gate drive. Four 10000 uF capacitors sit atop the power board.

Testing: A 1.5 kVA 24 V transformer with rectifier and filter is the test supply. That gives about 33 VDC out. I'm OK with that. On perfboard everything was good. I put about 1000 watts through it with incandescent lamps, heaters, and brushed motors. Now that I have assembled it with the PCBs, it isn't nearly as stable.

The Problem: A small 1/3 HP induction motor caused instantaneous FET explosion. After repairing that meltdown, a 250 watt lamp turning off and back on blew the low side FETs. My little USB scope isn't fast enough to see what is happening before the destruction. Destruction happens with and without RC snubbers installed.

Solutions: Increase gate resistor value? Use a negative gate bias supply? Schottky diodes from source to drain?

I just need some back-and-forth conversation with people who know more than I do. Thanks for everything!
 

michael8

Joined Jan 11, 2015
410
Some datasheets:

IR21844 600 V half-bridge gate driver IC with shutdown and programmable deadtime
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462533600a4015355c955e616d4

TLP351
Inverter for Air Conditioner IGBT/Power MOS FET Gate Drive Industrial Inverter
https://toshiba.semicon-storage.com/us/semiconductor/product/optoelectronics/detail.TLP351.html
https://toshiba.semicon-storage.com/info/docget.jsp?did=16841&prodName=TLP351

StrongIRFET IRFB7545PbF
https://www.infineon.com/dgdl/irfb7545pbf.pdf?fileId=5546d462533600a401535619e1ca1e66

Seems a bit wierd to use the IR21844 half-bridge driver with > 1A output
current to drive input LED of the TLP351 which has a max input current
of 20 mA.
 

Thread Starter

PapaFet

Joined May 15, 2022
6
We can't have a discussion without accurate Schematics and Spec-Sheets to look at.
Indeed. I am plumbing the depths to make sure that this is the place to have that discussion. I have tried to talk on, say, Electronics Stack Exchange and get no feedback beyond terse quips. Here is a schematic of the quarter bridge that everything else is a copy of:Schematic_PowerPico Quarter Bridge_2022-05-15 (1).png
The "Connect" flag is either power ground or the interface between the upper and lower FETs, depending on where in the H-bridge you look.

A shot in the dark. What provision did you make for deadtime in the switching waveforms?
Dead-time is controlled by a resistor on the IRF21844 half bridge drivers. It is approximately 500 ns currently.

I understand that this is a vague question with many possible answers. But I don't know enough about power electronics to know what questions I should be asking. But I also understand that really this must have been a topic addressed in days gone by and I'm hoping that someone will read this and say "Oh, well, I had that same problem. Have you looked at ________?"
 

Thread Starter

PapaFet

Joined May 15, 2022
6
Some datasheets:

IR21844 600 V half-bridge gate driver IC with shutdown and programmable deadtime
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462533600a4015355c955e616d4

TLP351
Inverter for Air Conditioner IGBT/Power MOS FET Gate Drive Industrial Inverter
https://toshiba.semicon-storage.com/us/semiconductor/product/optoelectronics/detail.TLP351.html
https://toshiba.semicon-storage.com/info/docget.jsp?did=16841&prodName=TLP351

StrongIRFET IRFB7545PbF
https://www.infineon.com/dgdl/irfb7545pbf.pdf?fileId=5546d462533600a401535619e1ca1e66

Seems a bit wierd to use the IR21844 half-bridge driver with > 1A output
current to drive input LED of the TLP351 which has a max input current
of 20 mA.
Thanks for gathering those together here! Yes, you are correct that it is definitely non-standard. The reason is because the control system is a done deal. It was designed by an Australian fellow who posted progress on a couple of forums. He eventually got it hammered out and I ended up with one of the PCBs. Originally the board used a bootstrap circuit to drive high side FETs. Over time and testing, I learned it was exceptionally easy to destroy the 21844 drivers and/or Nano uC. So I added some optical isolation between the control and power board and an isolated gate drive power supply for each high side quadrant.
 

LowQCab

Joined Nov 6, 2012
4,029
A search comes up blank for an IRF3905, is that a misprint on your Schematic ?
However, IRF4905 is a good number,
but it's only rated for -55 Volts, 74-Amps which is certainly not going to survive.

75-Volts is probably the bare minimum, 100-Volts would be better,
along with a stack of TVS-Diodes across the Transformer-Windings, ( Primary and Secondary Windings ).

You're almost quadrupling the Supply-Voltage,
so this means the Output-Current will be roughly ~1/4 of the Input-Current, or,
the Primary-Current will be ~4-times the Secondary-Current,
so if You attach a Load that pulls ~20-Amps @ 120-Volts,
even for a half-Second, ( Motor start-up ),
the FETs will see around ~80-Amps, or more,
which equals, instant release of the Blue-Smoke,
( in the case of a 74-Amp-rated Part ).

You'll need Current-Limiting if you're going to beat on it.

What's the actual FET number that You keep smoking ?

Where's the rest of your Schematic ?
.
.
.
 

Thread Starter

PapaFet

Joined May 15, 2022
6
A search comes up blank for an IRF3905, is that a misprint on your Schematic ?
However, IRF4905 is a good number,
but it's only rated for -55 Volts, 74-Amps which is certainly not going to survive.
Yes it is mislabeled on the schematic. Should read IRF3205. 55 volts, 110 amps. It is correct in the body of the OP.
75-Volts is probably the bare minimum, 100-Volts would be better,
along with a stack of TVS-Diodes across the Transformer-Windings, ( Primary and Secondary Windings ).

You're almost quadrupling the Supply-Voltage,
so this means the Output-Current will be roughly ~1/4 of the Input-Current, or,
the Primary-Current will be ~4-times the Secondary-Current,
so if You attach a Load that pulls ~20-Amps @ 120-Volts,
even for a half-Second, ( Motor start-up ),
the FETs will see around ~80-Amps, or more,
which equals, instant release of the Blue-Smoke,
( in the case of a 74-Amp-rated Part ).

You'll need Current-Limiting if you're going to beat on it.

What's the actual FET number that You keep smoking ?
I have killed both IRF3205 and IRFB7545 FETs.

Where's the rest of your Schematic ?
.
.
.
I didn't provide it because it's just more of the same. That driver and FET circuit repeated four or eight times, depending on whether I double up the FETs in parallel. I can certainly provide such a diagram if necessary.
 

michael8

Joined Jan 11, 2015
410
Dead-time is controlled by a resistor on the IRF21844 half bridge drivers. It is approximately 500 ns currently.

TLP351 (schematic has TLP350?) has a propagation delay time of from 100 nS to 700 nS. Since you have two
if they are different by more than 500 nS you can still get overlap. I'd guess that this isn't "the" problem because
there is likely to be more than one.

I'm guessing that this is a "build one" rather than build thouands so cost minimization isn't that important.
If so, I'd try to include some circuitry to monitor overcurrent (and possibly overvoltage?) and have it quickly
shut things down, hopefully before the fets explode. And it would need to latch the error condition so you
could figure out what happened when everything suddenly stopped (without visible or hopefully invisible damage).
 

Thread Starter

PapaFet

Joined May 15, 2022
6
Seems like the FET shorts out the 10000uF capacitor when it switches on. There can't be a good outcome to that!
You know what? I can see how that would be misleading. That is why a complete schematic is needed. I will amend this tomorrow and get a full schematic posted. Thanks for pointing that out!
 

atferrari

Joined Jan 6, 2004
4,764
Sorry to tell you but "conversation" smells too similar to blah, blah, blah.

In my job I use precise sentences, with the minimum of words. That way I convey the intended meaning, do not fatigue the reader and do not give way to unnecessary questions back to me.

Ch. Dickens wrote long texts but the pleasure of reading with him is granted from start to end. If you have time take the pain of reading "On writing" by S. King. His economy in the use of words is notorious.

Nobody likes long texts. Nobody.
 

Thread Starter

PapaFet

Joined May 15, 2022
6
Sorry to tell you but "conversation" smells too similar to blah, blah, blah.

In my job I use precise sentences, with the minimum of words. That way I convey the intended meaning, do not fatigue the reader and do not give way to unnecessary questions back to me.

Ch. Dickens wrote long texts but the pleasure of reading with him is granted from start to end. If you have time take the pain of reading "On writing" by S. King. His economy in the use of words is notorious.

Nobody likes long texts. Nobody.
Very eloquent. But I think we may be thinking of different definitions of conversation. You see, I am not an expert in this field and therefore I require a tad more than short equations between colleagues. Something like: "I have blown up a number of MOSFETs, mostly when I power an inductive load."

Someone with more experience: "Have you looked for ringing on the gate of the FETs?"

Sorry to have caused you to waste time getting on my level.
 

Solar Mike

Joined Apr 26, 2018
33
Issues I see in your (part Schematic).
Those TLP351's don't have sufficient drive current for large high input capacitance mosfets when switching at 20 odd KHz SPWM.
The mosfets need a higher voltage rating, 80 - 100 V, something like IRFP4468.
Those opto couplers are very "old" and slow, with a huge range in switching times between devices, If you are having multiple parallel mosfets driven, then the timing overlaps could be a serious problem, leading to cascade failure.

If you want isolation between your cpu and the mosfets, keeping the IRF21844 for dead time generation, I suggest you ditch the TLPXXX devices and go for something like UCC5350SB, with isolation, under voltage lockout, higher current drive and tight variation between devices. One of these could drive 2 large mosfets or a buffer pair of transistors driving 4 or more.
 
Top