IGBTs burned by an unkonwn bad spirit

I'm trying to do an inverter based on Sine Pulse Width Modulation (SPWM). I did two stages, control and power. The first, done with AmpOp 741 and comparators LM311 generates 4 signals wich would be injected on the gates of a bridge of IGBTs (G4PC50K IR 022P), these receive a DC signal to be modulated by them. The power schematic is attached. View attachment IRG4PC50.PDF
I made a record of the signals out of the control circuit and used the data in excel. The result is that the signal es zero during certain time, then describes a sinuidal figure on positive side, wich is the expected signal. They are not in phase with the oposite branch, so it doesn't go to short circuit, so I suppose they are right.

Anyway, the first test was to feed just the power stage, with 110 V to the bridge rectifier wich would feed the IGBTs with a rippled DC (it has not a capacitor yet), but, with the IGBTs off (no signal in the gate), I supposed no current would pass trhough. So I would have just a rectified signal feeding IGBTs and nothing out. But the result instead was a slpike and that the IGBTs 2 and 3 was burned.

The first option I think is that the IGBTs do conduct without a signal in their gate (on air), so if they are not connected to ground, they will conduct, so they go on short circuit inmediatly.

The second option is that I have something wrong in the design, but I don't see it. (see schematic)
The design is quite simple, and I have checked the circuit several times so there are not short circuits between traces and obviously that heat sinkers are correctly isolated.
Indeed the first spike was a trace, so I thought could have been a fault on isolation (traces too near), so I fixed it and then the IGBTs were blowed.

The third option, as other thread says, could be the absence of the capacitor on the rectifier (wich I thought would give us a rippled signal but DC at least), wich could bring an overvoltage situation on the IGBTs caused by inductance, although it would remain anyway the question about if the IGBT conducts with the gate 'on air'. It is not clear to me what capacitance is right for this circuit. I was thinking on a microwave capacitor, of 1 uF 2000 V.

I have to know what happened and solve it before replacing any component, so please, if somebody knows it, I'll be very grateful with your advice.

When it be resolved, I have a further question about the capacity of a SPWM inversor or inverter to drive an induction universal motor, and if it's velocity can be varied with a third stage circuit based on a triac with zero passing detection circuit, although the frequency of the control (10 KHz) be higher than what can drive the moc3021 and the Triac itself. That is, if it can drive velocity between 80 to 100% although it 'jumps' steps of the spwm.



Thank you for your time
Luis Jiménez I.



- - - Updated - - -

Note that it has no charge at all, it is an open circuit in this sense.

- - - Updated - - -

Note that it has no charge at all, it is an open circuit in this sense.

Yes, option 1 is a sufficient condition to burn the IGBTs. Never power an IGBT or MOSFET with open gate.

Option 3 can work too, but not without a bit of load inductance.

Hi,

Your schematic is confusing. Please avoid unnecessary lines, avoid unnecessary signal crossings, avoid thick lines that hide the junction dots.

It seems you have a single sided PCB ... this usually is not suitable for switching power applications.
Aditionally I miss fast capacitors at the power supply and maybe you need fast diodes across the IGBTs.

Klaus

FvM: Thank you very much. I'm in debt with you and the forum, because with this fault I would have burned several other IGBTs.
I rated the IGBTs for its current, 70 amps for a max peak load of 30 amps., but do you think that considering voltage this IGBT is (G4PC50K IR 022P) ellegible for a 220 V inverter? As I will have to buy them again, it is important to know if they are the correct selection.

- - - Updated - - -

where or how can I select and find this fast capacitors and diodes? I buy on mouser or arrow, but what condition most I look for to set these conditions? So there will be back current if I don't install these diodes? What capacitance would be right? I suppose it is to prevent inestability but i don't know where should I put them. Consider I'd like to handle 220 V.

your bride rectifier will likely get too hot and die without heatsinks, the gate driv should hold the IGBT's firmly off whn they are not supposed to be on - even at power up

you need largish caps right on the HVDC near to the igbt's to avoid over volts at device turn off, if the igbt's do not have internal anti-parallel diodes then you will have to add then externally...

It is very hard to make a success of this type of circuit without some experience in building power supplies and inverters... circuit layout, heatsinking, RFI reduction, properly rated parts, clean gate drive - all are necessary ....

Use 10k ohm to gate to bleed floating gate charge with no drive connected.

Control timing of Gates to have a dead time of at least 300ns or > L/R of load to prevent "Shoot-thru short circuit."

Hi,

There are so many proven circuits around, many datasheets and application notes, many dicussions...
A forum can't replace school ...

I recommend to go to an IGBT manufacturer's internet site, look for datasheets and application notes.
You can trust them, they are very detailed and complete.
Don't trust every (simple) schematic you find by a google search.

Klaus

True and wise words have been spoken

This time, the forum didn't replaced school but replaced the teacher to which one ask. I studied phisics and digital machines: logic gates, carnaugh maps, microprocessors, programming, but respect analogic circuits I had big lagoons. Nevertheless, studying different texts thoroughly, papers, simulating with proteus and with some advices from the forum I have now a working spwm control. I recorded 300,000 pairs of data with a card osciloscope, passed them to excel, made a filter to know when the branch is on, I did a class separation, added the pulses and got graphics about the wave form I get from this circuit (because in the osciloscope is not possible to see the out wave form at 60 Hz if it is spwm over a 10,000 Hz signal), confirming it works properly.
Now, with the power stage, I didn't know they would burn if the gate were on air. I'm now collectig information from different providers, including IGBTs modules, studying about heat sinks, etc. In a pair of weeks I'll have a functional inverter with the specifications I want, and later I'll do another but PIC based. I found several mistakes in an university paper about inverters, but I had to do it to know what was wrong. If somebody insist enough, it is possible to make the puzzle, but the forum saves tons of hours and frustration trying to learn what others know already. When I can get some money from this, I'd like to share some with the ones who want it.
Thank you for all your patience and helping words, which reconciliate us with the idea of a human brotherhood like Lenon used to say. This is the control circuit I'm talking about.

- - - Updated - - -

About the proven circuits you mention, on internet it is a luck to chose which is really working. Is it in somewhere a repository of proven circuits on which it be possible to trust? As I told you, the basic concepts of the paper respect spwm, was right, but there were several mistakes in the circuits proposed.

Trust data sheets or app notes from reputable manufacturers.
They are almost always correct, although there have been exceptions, which are corrected on the next revision.

Hi,

YES. As already said in post#7....

The semiconductor manufacturers are the source of information you can trust.
They can only exist if they sell their parts. And they can only sell their parts if the circuit designers know how to use them.
Thus the manufacturers will assist the designers by providing good and proven material like datasheets, application notes, simulators, development boards...

In opposite... if they provide no, no good or erroneous information .... no one will build a working circuit and the manufacturers won't earn money...

Klaus

Now another thing. I'm securing that the control circuit doesn't emit for any reason a short circuit state. Then next question is if IGBTs gate support to be feeded directly by a CMOS gate who emit 12 V, without amplification with a transistor? The gate datasheet says it can support just 2.4 mA. The IGBT datasheet says 5 volts are enough to set on the IGBT but it doesn't say specifically what current it needs. What do you think?

Hi,

driving an IGBT gate with just 2.4mA .. at a swtiching frequency of 10kHz.

You surely need to read application not on "how to drive IGBTs".
The probelm is not the DC current into the gate..this may be just several microamperes.
The problem is the switching edge. There are unwanted capacitors inside the IGBT and there additionally is a so called "miller effect". You need to charge/discharge the capacitors fast enough, else you get bad efficiency and bad output waveforms.

There are dedicated IGBT driver ICs. Usually they can source/sink several 100mA to several Amperes.

I assume every IGBT circuit with a switching frequency in the 10kHz range will show the driver circuit.

Klaus

Like a Mosfet, the gate of an IGBT is a capacitor which is charged and discharged each time you turn it on and off.

The average current is I = Cg * Vg * f.
Calculate it....you may see that depending of the switching frequency, the average current can be hundreds of milliamps.
In top of that, the peak charging required for fast transitions.

Use a proper IGBT or Mosfet driver. Most semi companies manufacture ICs specifically for this task.