# RC snubber network not protecting IGBT

I'm working on an electromagnetic accelerator project but I'm having trouble protecting my IGBTs that switch current through the coils. Here's the schematic:

C1 gets charged to around 250 V (In my tests I only charged to 200 V) and is then placed in series with L1 when Q1 turns on. L1 has a resistance of 0.7 Ω so we expect 285 A when running at 200 V. So the energy in L1 is around 12 J. Q1 is an IXXH80N65B4 which is rated for 430 A surge an 650 V collector to emitter. So C2 has to hold 12 J of energy and not exceed 650 V which means C2 must be at least 56 μF and R1 must be at least 2.3 Ω.

I didn't have this size capacitor on hand, the next closest I had was 1 μF, but I wasn't too worried about this because I thought most of the energy would be dissipated by D1 and not much would be absorbed by the snubber. So I tested 1 μF with 4 Ω and this destroyed the IGBT and shorted all three pins.

I assumed 1 μF was too small, so I decided to try a much bigger capacitor to hopefully reduce the voltage spike since $$\V=\sqrt{\frac{2E}{C}}\$$. So I used a 1000 μF capacitor with the same 4 Ω resistor and yet again destroyed another IGBT. Not only that, but it also shorted the 1000 μF capacitor which was rated for 450 V. I don't know why this snubber isn't working. I thought it was safer to use bigger capacitors to limit the voltage?

Simulation Results:

After simulating the circuit again the results are not the same as I remember (C2 is parasitic capacitance of IGBT between collector and emitter).

Here's the collector voltage

However, it seems LTSPICE thinks the snubber does not make a difference as when I remove R1 and C1 the plot remains the same. However, when I remove D1 we get this:

Which doesn't look too good! (The plot above is with 1000uF snubbing capacitor. If I use 1uF its even worse at 1.8kV peak). So could my diode be the issue and if so do high surge current schottky diodes exist? Or could I try to reduce this spike with just a snubber and no flyback diode

UPDATE:

So I tried this again but this time with an improved circuit to hopefully protect my IGBTs. Here's the circuit:

Changes include using a 1.2$$\\Omega\$$ resistor and 150uF capacitor for snubber as well as gate filtering and protection. I tested this circuit and it worked for 60V and 100V on C1. But then I tried 160V and the IGBT failed once again. However, I did not measure any overvoltage at the collector on the oscilloscope (The trigger level was set to the same voltage as the capacitor, taking into account the voltage divider, and it did not trigger meaning the voltage stayed below the capacitor voltage which is what simulations predicted). This has left me very confused as there is not over voltage on collector and I know the current through the IGBT is definetly within the IGBT's rating. I have more IGBTs to use but I don't want to destroy more. I have no idea what is wrong.

Here's a photo of the setup:

And here's a full schematic if it helps:

Another Idea:

I realised that my IGBTs are only specified for 430A for 1ms and 160A continuous in the datasheet. So depending on how fast the projectile moves it may need to sustain current larger than 160A for a few ms. I've seen others use IGBTs which are rated for 800A for 1ms instead and have had success running them at 400A for around 7ms. Is there any other safer option to get around this problem rather than getting even more expensive transistors and just hoping they can handle high current for more than 1ms?

UPDATE with new transistors:

So from my testing I thought excessive collector current was destroying my IGBTs so I got some 160N65B4 IGBT which can do 860A for 1ms. So I ran the circuit again and what do you know it breaks yet again. At this point I'm pretty lost on how to move forward, even if my flyback diode is too slow the snubber should control the collector voltage on it's own (assuming the transistor is on for 4ms or more which is reasonable on time from what I've researched). I checked over the circuit and moved components to reduce stray inductances. This included moving the flyback diode physically closer to the collector of the IGBT. However, the ground wire which connects the ground of the IGBT board to the battery ground is quite long and has around 2uH of inductance. I'm unsure how big of an issue this is but it does concern me given that any small inductance could be a big problem due to the high operating current. Should I be worried about this inductance? If not I'm all out of ideas.

• What are the specs on the diode? It may need to be a fast switching type. And perhaps use a snubber capacitor closer to your calculated 56 uF. Make sure it has a high surge current capacity, and a low inductance resistor. Have you tried simulating this? Commented Feb 6, 2023 at 5:30
• Diode is a GI754-E3/73 rated for 400V and 400A surge with 2.5us reverse recovery. I did simulate this a while ago but didn't save the file. I'll simulate again just to check Commented Feb 6, 2023 at 6:02
• 2.5 us is pretty slow. Use an actual diode model in the simulation. You should be able to use several in series to get adequate voltage rating. 400V may not be enough. Commented Feb 6, 2023 at 6:58
• I think you should simulate the circuit to include the application of the 200V supply, and the turn-on and turn-off of the IGBT. And I would expect to see some ringing at the transitions, due to intrinsic capacitance and inductance of the components. There may be clues to the failure with an expanded view of these events. Commented Feb 6, 2023 at 8:09
• The R+C is more-or-less irrelevant with the diode present. Please show layout: you may be subject to stray inductance in the gate or collector loops. Commented Feb 6, 2023 at 11:35

1000 uF for C1 is a bit too big.

Here is a simulation for only the snubber and then diodes were added.
Diodes must sustain a high current for a "big" time (at least 1 ms) ...

And the power sustained by the diode (Peak and Averaged).

• Confirmed. No change in simulations. Commented Feb 7, 2023 at 12:30
• Ok I've got some new parts coming including a smaller 150uF capacitor to try and a 1$\Omega$ resistor. Still not sure why that 1000uF didn't work but I found out that I didn't actually short that capacitor as I was unknowningly measuring the ESR. This means somehow the collector voltage rose above 650V even though the 1.2$\Omega$ resistor should've limited this voltage to well below that. Also, should we add a diode between emitter and collector to limit negative voltage spikes since the IGBT I'm using does not have a body diode. Commented Feb 7, 2023 at 21:32
• Interesting thing ... I have noted (in simulation) that if L1/r1 = r2*C2, there is no overvoltage on the IGBT "collector". I choose C2 = 600 uF and r2 = 0.7 Ohm (for L1 = 290 uH and r1 = 0.7 Ohm). Commented Feb 8, 2023 at 20:50
• Ah ok. I must've had a mis-connection or cold solder joint because all the simulations seem to suggest that the back emf should be handled. I've made some additions to the circuit including some gate filtering and protection as well as a voltage divider between the collector and ground so I can scope the voltage with my oscilloscope without running the risk of destroying it as its only rated for 400V. Hopefully being able to see the collector voltage should help diagnose the issue. Commented Feb 9, 2023 at 2:31
• A try. I should not use r2 and c2. Commented Feb 10, 2023 at 11:44

I finally have discovered the issue. After 6 dead IGBTs I asked for help from someone I knew who had his own accelerator. He advised me to add a resistor in series with the flyback diode to limit the di/dt through the coil and thus limit the inductive kickback voltage. So I added a 10ohm 10W resistor. After making this change I did a shot at 150V which would've killed the IGBT in my original circuit. It fired and my hands were shaking as I went to check for a short on the IGBT. And you have no idea how relieved I was to not hear that beep. It finally works! So I guess the current through the flyback diode/coil when the IGBT turned off was so large and quick that the di/dt was super big and created a large spike. And of course the snubber has no way of limiting this spike caused by the diode/coil. Thank you everyone for sharing ideas and providing help.