Gate and Collector of IGBT is getting short

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
Hello all,
Good day!
I am new to this forum thus please pardon me if I make any mistake while asking my question.

I have been trying to make a small VFD (variable frequency drive).
I have used STGP19NC60KD IGBTs for the purpose and using IRS2109S for driving the gate.
I am driving the IGBT with a PWM frequency of 8KHz.
It works well when no load is connected and I can see the required output.
When I connected a motor, the motor ran for 5 seconds, and then the first set of IGBT(high & low side) which is used for driving phase R got burnt, and its driver IC is also damaged.
When I checked at IGBT terminals, I see that its gate and collector is short-circuited.
I replaced the IGBT and again tried, but it still burnt the same IGBT.

In first attempt, it burnt both high and low side IGBTs and all the resistors and zener diode.
In the second attempt, it burnt High side IGBT only and driver IC.
Also all the resistor on both high/low sides and zener diode were also damaged.

I have tried searching a lot but couldn't find anyone with the same problem.
I see that Collector-Emitter or Drain-Source short circuit is a common problem, but in my case, gate-Collector is short circuit and I can't find any explanation for it.
I have also attached the schematic of one of the IGBT path.

I really appreciate any help to show me the direction of the problem.
Thanking You all.

Sch.png
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
Do you have any snubber circuit or catch diodes around the IGBTs?
Hi Albert,
Thanks for the reply.

No, I do not have any other circuit other than I have attached in the image.
This is exactly the same circuit I am using.

The circuit is powered directly from 230VAC and is rectified to DC. The Bus voltage is around 340V DC.
 

carloc

Joined Oct 8, 2018
13
Collector to gate shortcircuit is not an uncommon failure mode.

Freewheeling diodes are luckly co-packed with the IGBTs themselves, so no worry on this side

upload_2019-9-2_19-7-3.png

I could guess some stray inductance problem.

The point is that the two IGBTs shall be connected to the power supply by very short, large connections.
Power supply itself shall be well bypassed by some low stray inductance capacitor very close to the IGBTs.

DC bus schematics , pcb layout and photos would also help a lot.

This stray inductance can give two similar issues;
  1. Vce overvoltage peak at turn-off possibly exeeding BVces and hence frying the poor IGBT
  2. The same inductance can resonate with stray capacitance causing Vce ringing, bootstrap gate drivers like the one you are using are sometimes very sensitive to fail if high-side driver emitter/source connection goes negative w.r.t. ground.
Upon gate driver failure you loose IGBTs control which subsequently fail.

In both cases above (and anything else too) a couple of scope screenshots would invaluable.
Please, for your safety, use only isolated differential voltage probes.
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
Hello Carloc,

Here I am attaching some of my PCB photos.
I checked it today as well by replacing with different mosfets but still it got burnt.
I observed that only R phase IGBTs are burning and other are working perfectly.
This R phase IGBTs are at the farthest distance from the output terminal, can it be a problem?

Please see the attached images of my PCB and layout.
I have a USB oscilloscope and I am really afraid in this situation to connect my laptop to the VFD via my USB oscilloscope. Thus I can't attach scope waveforms.

@Alec_t , I have attached the schematic only for 1 phase, rest are exactly the same.
 

Attachments

Alec_t

Joined Sep 17, 2013
14,280
This is a long-shot. I see from the Mosfet-arrangement pic that the trace going to the gate of T1 (in the R phase) runs more or less parallel to the return ground path for a significant distance. Is it possible that large current pulses in the ground path cause noise pulses to be inductively-coupled into the T1 gate signal and result in false triggering of T1?
 

carloc

Joined Oct 8, 2018
13
No the phase connections are the only non critical ones, they are anyway connected to the motor, plenty of inductance over there.

But beside that, unfortunately your pcb doesn't look so good as far as grounding and power connections are concerned.
Main power supply makes large loops and lacks of bypass capacitors,
Board.png

Also gate drivers and general board grounding looks weird, probably one or more ground planes could have served better.
Just about any connection should be thought considering its inductance and the current (the AC one, di/dt actually) passing.

I believe you could try slowing down switching or adding a few nanofarad decoupling capacitors on DC bus, very close to IGBTs themselves but the truth is that is just blind shooting.

I am afraid you should really get a suitable differential probe and eventually redesign pcb layout to have a reliable VFD.
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
Hi,
Thanks for the detailing.
I understand I did the AC side routing very causally. This is my first high power electrical design and I had been expecting a revision in this.
But I thought this would have been Ok for the initial prototype tests.

In this case, my motor runs well for some time and IGBT fails after a few seconds.
We had a 5Amp fast blow fuse, Circuit breaker in line for protection and they never trip. We had connected a multimeter in neutral line for measuring the return current and it never crossed 1.8Amp during motor run. It showed 2.8Amp only when IGBT bursts.
Also, I checked for the shorts on heatsink before connecting voltage to the system and there is no short through it.

I will try by reducing the switching frequency. Now it is 8KHz, I will try by making it 4KHz.
Also, I would like to give it a try by adding some bypass capacitors directly on IGBT.
Is it ok if I add 100nF/1KV between collector and ground for the High side IGBT?
Should I also add 100nF/1KV between collector and ground for low side IGBT?
There is a bootstrap circuit for high side IGBT, can these capacitors have any effect on it?

With the current design I hope to get at least the motor running somehow, and I will redesign the PCB by removing all these problems.
Please suggest. Thanks for your help.
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
This is a long-shot. I see from the Mosfet-arrangement pic that the trace going to the gate of T1 (in the R phase) runs more or less parallel to the return ground path for a significant distance. Is it possible that large current pulses in the ground path cause noise pulses to be inductively-coupled into the T1 gate signal and result in false triggering of T1?
Sorry, I can't see them parallel, at the max they are slightly angled to the return ground path.
Also, as we checked the current by connecting multimeter in neutral, it never crossed 1.8 Amp during the motor run, and it was 2.8Amp during IGBT failure.

Also, the fact that only R-phase IGBTs are getting fried, not the others while their trace to the gate are more closer and angled to the return ground.
 
Last edited:

shortbus

Joined Sep 30, 2009
10,045
Also, the fact that only R-phase IGBTs are getting fried, not the others while their trace to the gate are more closer and angled to the return ground.
My thoughts only on this. What do you know about the motor? Could it be damaged? What happens if you move the "R" connection to another phase in the motor? Does the IGBT blowing follow that move? I know doing the last two will change the rotation direction but will determine whether the problem is with the board or the motor it's self.
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
My thoughts only on this. What do you know about the motor? Could it be damaged? What happens if you move the "R" connection to another phase in the motor? Does the IGBT blowing follow that move? I know doing the last two will change the rotation direction but will determine whether the problem is with the board or the motor it's self.

No, motor is perfectly fine. It is a brand new motor and runs good with another drive. Also their isn't R, Y, B markings on motor and thus i always connect motor wires randomly to the pcb.
 

carloc

Joined Oct 8, 2018
13
Well, as already said without any scope signals is really hard.

When I said you could try slowing down switching I meant rise and fall times, i.e. increase gate resistors, not swtiching frequency.

If they get fried for overvoltage adding 100nF between highside collector and its lowside emitter may help.
If you have some ringing around, gate drivers' GND it's going to be harder.

Good luck.
 

Thread Starter

Krunal bhongade

Joined May 28, 2017
7
Hello,
Thank you for your inputs.

According to the discussion above, I have redesigned my PCB and have made the following changes.
The schematic is same as attached in my first post, except I have added capacitors at the Drain of mosfet as discussed above.

1. Used 800V MOSFET of TO247 package (bigger in size and more space between 2 adjacent pins), also the PCB will be milled in between the gate and drain pins. Used MOSFET: IPW80R360P7XKSA1
2. Added 100nF/1KV capacitors between high-side collector and its low-side emitter.
3. Modified layout and added ground planes, please check the attached image.
4. Heatsink is placed in the middle of PCB, AC input and output is kept above heatsink and low voltage side is kept below heatsink.

I feel that it would work this time but I need an expert eye just to have an overlook and add their comments on it.
I am sure the discussion will be very helpful even for others having the same issue in the future.
Thank You all.
 

Attachments

Top