Has Anyone Tried IGBT IXGX72N60C3H1?

I'm not sure why you picked those parts, they're expensive ($11) and seem not to be stocked anywhere. IXYS says that these part's claim to fame is ultra-low V_CE(sat), 1.35V at the rated current of 72A, and that they're intended for use up to 5kHz. So they have far higher current capability than you need, and are slower than you might want. You know, C_ISS = 6600pF, Q_GC = 80nC, etc.

In general I can recommend IXYS products to you, having used lots of different parts of theirs and having lusted after many others. They often push the envelope in a gratifying way. But for your application I'd think you'd want a fast, reliable easy-to-get jellybean part. For example, we might pick '12N60 parts rather than '72N60. These are plentiful in stock as HGTP12N60A4D from Fairchild and IXGP12N60C from IXYS, are rated at 600V, come in convenient appropriate TO-220 packages, and cost $3.07 and $3.69. They're intended for high-frequency use, to beyond 100kHz. The IXYS part has V_CE(sat) about 1.7V at say 5A, and C_ISS = 860pF, Q_GC = 10nC. They claim a 55ns switching time. In general the IXYS parts looks better than the Fairchild part, which has much higher capacitances, etc., but I'm sure they'd both work well for you.

At your modest power levels you might consider a MOSFET instead. As long as we're talking '12N60, how about a FQP12N60C from Fairchild? Or get a smaller part closer to your needs, like one of Fairchild's new '7N60 offerings. They have the FCPF7N60N with R_DS(on) = 0.46Ω and Q_GD = 6nC, $2.24 at Mouser, which looks good. Or their FDPF7N60NZ with lower charge but higher R_DS(on), for $1.40 each. Nah, I like the first one better. It would allow you to really speed up your switching frequency and reduce the size of your magnetics. OTOH, if your frequency is dictated by other considerations, such as an ultrasonic transducer resonance, then this possibility may not be appealing. But the MOSFET would still appear to offer lower switching losses than a IGBT.

I do not know where things were wrong but MOSFET of IXYS 1kV 12A version were blown in few seconds.

I am now looking at higher power dissipation factor IGBTs with PTC nature. I have selected one IOR part RG7PH42UD1PBF 1200V 30A 313W/125W @100oC. These are somewhat expensive but will get these some 500 numbers soon for trials.

IXYS parts will come in December as there i nearly 4 months lead time.

I am going to stay with IGBTs for 20kHz to 30kHz range and 500W to 1000W power switching. I am refining my design for better transformer function, some kind of voltage regulation and proper EMI filters.

My driver was drawing in excess of 10A in primary for 2A 1kV in secondary. Push-pull input at about 200V and input current should have been only 5A but it was in excess o 10A so I need find out where things were wrong. Power loss is a big problem.

Thanks for writing.

Hah! Well, they probably weren't burning our because of an insufficient V_CES voltage rating! But you do have to watch out for reverse V_CE, which can quickly destroy the part. That's why many IGBTs have an added reverse diode taking up valuable space on the die. You didn't mention the p/n of your blown IGBTs, but I wonder if they had diodes?

Of the IGBTs I suggested, the IXYS IXGP12N60C does not have a reverse diode whereas the Fairchild HGTP12N60A4D does, and of course MOSFETs always have one.

And for the IRG7PH42U you're considering, the regular one does not have a diode and the one with a D1 suffix does. But again, those parts are massive overkill, with their 85 to 90A 25°C rating and very high capacitances and charge. Those two issues can cause their own problems.

Back to the diode story. At any rate, it's easy to add your own reverse diode across the part. Actually, many engineers prefer to use their own soft-recovery diode, rather than live with the inferior internal diode provided by the manufacturer. Recovery-time snap-off can make serious short 5ns-long spikes, etc., that will destroy sensitive IGBT gate oxides.

As I implied, you may have a common inductive-spike gate-damage problem. The way to debug these is first, with your small appropriate intended part, run it at reduced voltages and currents, and carefully probe for excessive spike voltages anywhere, such as flyback from a diode recovery-time current-snapoff. You look for a reverse diode turning on, then search out the point where it turns off, and expand the horizontal scale at that point. You may have to set the scope to trigger on the spike to see it. You can set the trigger to arm just before the suspected event. One other thing, the inductance in your scope probe's ground lead may obscure the real spike waveforms.

Once you find the culprit, you can gradually turn up the voltage and current and see how it progresses, and discover what you need to work on, without destroying the part and the evidence. For example, you may need improve your snubber network. Don't have a snubber network? Aha, maybe that's the problem! A proper snubber can prevent reverse diodes from conducting in the first place, thereby preventing any snapoff spikes.

I have all of them with diode. I will try half bridge design again to see how this works. I am also going to add PFC to make voltage constant near 400V. I will give you feedback.