Hello. I'm working on a single IGBT module circuit with high side switching. This is for a plastic sealing application (apply controlled power to a heat element wire). My circuit works well in simulation (LTspice) and I've built the circuit and tested successfully at lower power levels. When high power is applied the IGBT is destroyed immediately on the first gate activation (holds voltage fine until fired). I have read several application notes and studied the specs and I'm pulling my hair out.
Background: This control is typically done with SCR or triac based phase angle control. I wanted to test an IGBT based design because of their availability and high current capability. The goal was to simplify the circuit and minimize components so that our field techs can troubleshoot. I also wanted to minimize changes to our overall machine design, which is why I chose to control the secondary side of our seal transformer (not shown in the schematic, 3 phase 460VAC to single phase 160VAC). Feel free to comment on the feasibility of the IGBT vs SCR methodology, but the purpose of my post is to learn why my IGBT is failing at higher currents. It seems to me that I'm well within the safe operating range of the device. I have a large heat sink attached, but this doesn't seem to be a thermal issue (fails immediately, never gets warm). Here's the details:
IGBT model: IXYS IXGN320N60A3 (data sheet attached)
DC to DC converter for the 15VDC boost (shown in the schematic) is done with a Recom Power RK-2415S/H (4kvdc isolation).
Benchtop test circuit (works well) - 120VAC supply, rectified to 195VDC peak to peak (oscope), 1khz pwm signal to a mosfet gate. The mosfet controls voltage to the IGBT gate. Load is a 60ohm incandescent lamp. Output waveform is clean. Test successful.
Simulation in LTspice - Schematic and results attached. Note the "DC" source to the IGBT is rectified but not filtered or smoothed. Smoothing the output is challenging due to the rapid discharge to the 1ohm heater element (the capacitor would be huge). I decided to eliminate smoothing capacitors because the load in this application doesn't care about the waveform. Test successful.
High power test - 1 phase 160VAC supply from a transformer, rectified to 210VDC (meter reading, likely 227 peak), same gate control circuit as used on the bench test. Load is a 1 ohm heater wire. The seal power via PWM was set to 5% and the seal time was limited to 500ms. Voltage is held in the off-state with no problems. IGBT immediately fails (audible pop, bright blue arc flash) when the seal is triggered. At the 227VDC on a 1 ohm load, current should be 227A. My IGBT has a max Ic of 320A and Vce of 600V. I'm at room temp. Per the manufacturer's safe operating parameters, this should be okay. What am I missing?
Please note that the circuit shown is for proof of concept only and is not intended to be a robust final design (be nice if you can). I have not included a snubber circuit because the load is not inductive and, again, proof of concept only. Also, this is not my engineering area of expertise, so roast me if its warranted.
For information, our traditional seal unit (triac phase angle based) uses about 70VAC @ 70A to seal plastic. FWIW, I now believe the phase angle method is more robust, but I still want to understand this application.
I'm sure I've left out something so please just ask and I'll add detail as needed.
Thank you very much for any suggestions or thoughts you can offer!
--- Updated ---
IGBT datasheet
--- Updated ---
I see that my IGBT has a "terminal limit" of 200A, so perhaps I'm arcing across terminals. I am crossing that threshold. Would that short the IGBT? Perhaps this is the answer but I'd like to hear other opinions. If the IGBT just needs to be bigger, that's fine. I just don't want destroying them to be my test method. Thank you.
According to your schematic, you are switching IGBT on with 1k series resistor. Typical gate resistor would be factor 50 to 200 lower. The high gate driver impedance causes very slow turn-on with high losses and possibly self-oscillations during turn-on. You should definitely use push-pull gate driver with lower impedance.
According to your schematic, you are switching IGBT on with 1k series resistor. Typical gate resistor would be factor 50 to 200 lower. The high gate driver impedance causes very slow turn-on with high losses and possibly self-oscillations during turn-on. You should definitely use push-pull gate driver with lower impedance.
You're correct about the slow turn-on. I found it to be very evident on the scope with gate resistors at or above 5Kohm. I'll adjust to a lower value. I'll also research the push pull gate drive suggestion. Will that work with high side switching? Thank you for your help.
Are you sure your simulation schematic is correct?
V2 is referenced to the IGBT emitter, but M1 pulls the gate down to GND when turned on. If M1 turns on fast while the AC line is high, then the gate voltage may drop faster than the IGBT actually has turn turn off and the emitter voltage falls. This would lead to Vge going far negative, likely destroying the IGBT immediately. Also this would put a huge peak stress on R5.
Also if V3 ever stops functioning (loose wire, power loss, or some other failure), then U1 is going to turn on. Seems like a poor choice, especially if this thing is going to be delivering tens of kW.
I believe so. I'm a PLC/controls guy. I wouldn't even claim hobby level on this type of electronic design. This is the result LTSpice gave, but I can't say with high confidence that I have everything set up correctly within the software.
I did alter the name of the IGBT to match the model I used on the real circuit, not to mislead, just to prevent confusion about what IGBT I used in testing. I've actually used a few different models of IGBT but I don't think its relevant to the issue... just being fully transparent.
Perhaps the simulation sample rate is too slow and makes this look cleaner than it should be.
V2 is referenced to the IGBT emitter, but M1 pulls the gate down to GND when turned on. If M1 turns on fast while the AC line is high, then the gate voltage may drop faster than the IGBT actually has turn turn off and the emitter voltage falls. This would lead to Vge going far negative, likely destroying the IGBT immediately. Also this would put a huge peak stress on R5.
Thank you for this! This is exactly the conversation I was hoping to have. I've thought about this, but I wasn't sure which device would win the "race". After I read your comment I remembered seeing negative spikes on the oscilloscope during the bench test. I'm attaching that image from the scope (PWM at near 100%). R5 was likely 10K during this test so the turn on is very slow, and I think that makes the negative spike more evident. Is what's shown in this image evidence of what you've described? Excessive negative Vge? It looks like about 15 to 20V in the negative direction at each turn-off. So this would mean I was on the edge during the bench test, and of course destroyed it on the higher voltage machine test. I believe you've answered the "why?" in my question. Do you agree based on the image?
Also if V3 ever stops functioning (loose wire, power loss, or some other failure), then U1 is going to turn on. Seems like a poor choice, especially if this thing is going to be delivering tens of kW.
I agree. This is bad. My original plan was to cheat the gate drive circuit, so that I could prove out the high voltage/current portion. If the IGBT held up well with this pitiful gate drive then I planned to go back and use a gate driver IC or better design. When I realized that high side switching was required for safety aspects on the machine, my cheapo gate drive swerved in to strange territory. I became a little obsessed with trying to understand why I couldn't get away with it though, especially with that nice simulation taunting me.
FWIW, I did use a NC relay to ground the IGBT gate if any failure occurred in my microcontroller drive to M1. That wouldn't help a wire break or similar disconnection though.
I greatly appreciate the input! Thank you very much for your help!
After I read your comment I remembered seeing negative spikes on the oscilloscope during the bench test. I'm attaching that image from the scope (PWM at near 100%). R5 was likely 10K during this test so the turn on is very slow, and I think that makes the negative spike more evident. Is what's shown in this image evidence of what you've described? Excessive negative Vge? It looks like about 15 to 20V in the negative direction at each turn-off. So this would mean I was on the edge during the bench test, and of course destroyed it on the higher voltage machine test. I believe you've answered the "why?" in my question. Do you agree based on the image?
I'm guessing the waveform shows gate voltage? Can't tell what Vge looks like unless emitter voltage is also measured at the same time (and subtracted from the gate voltage, maybe with the scope's math function. Or by using a differential probe).
I agree. This is bad. My original plan was to cheat the gate drive circuit, so that I could prove out the high voltage/current portion. If the IGBT held up well with this pitiful gate drive then I planned to go back and use a gate driver IC or better design. When I realized that high side switching was required for safety aspects on the machine, my cheapo gate drive swerved in to strange territory.
I was going to ask why you chose to drive from the high side. I'm guessing there are valid safety-related reasons to make the load earth-referenced.
Aside from the potential issue with Vge exceeding its limits, FvM's point about the slow turn-on is also a potential explanation for the IGBT getting destroyed. Depends on the specific IGBT used and its SOA specifications...
As Easy Peasy said above, a proper gate driver (referenced to the emitter, with drive resistance in the tens of ohms) is likely worth investing in.