IGBT keep failing with mains, but rated at 600V ???

geekoftheweek

Joined Oct 6, 2013
1,201
I have a thought in the event this does turn out to be an over current situation. Inevitably either 2 days from now or 10 years down the road it will happen again. How about setting up a comparator to either short your input of your gate driver to ground (with the appropriate components), or use the comparator of the PIC to shut off the gate driver in software? I started playing with an 18f45q10 that has some configurable logic in it that isn't dependent on the program other than to set it up that would actually work out nice in this case.

I kind of forgot to mention I was referring to comparing the output of your ACS722 to a whatever voltage works out for say 10 amps or whatever.
 

MisterBill2

Joined Jan 23, 2018
18,179
I have a thought in the event this does turn out to be an over current situation. Inevitably either 2 days from now or 10 years down the road it will happen again. How about setting up a comparator to either short your input of your gate driver to ground (with the appropriate components), or use the comparator of the PIC to shut off the gate driver in software? I started playing with an 18f45q10 that has some configurable logic in it that isn't dependent on the program other than to set it up that would actually work out nice in this case.

I kind of forgot to mention I was referring to comparing the output of your ACS722 to a whatever voltage works out for say 10 amps or whatever.
Rather than a slow comparator to protect the base from over voltage, how about shunt diodes, arranged so that they will go into conduction as the base to emitter voltage approaches the limit? Diode clamping can be a very effective means of protection. And it is simple to verify experimentally, as well. In addition, a shunt diode takes no extra power, it is functional even when everything else is switched off. BUT you do need to know the base voltage limit value.
 

Thread Starter

Robin66

Joined Jan 5, 2016
275
Update: I measured the cold resistance of the lamp: 10ohm. This would result in a potential max current of 33A on cold start up, depending on the AC phase angle when I happen to flick the switch. This exceeds the abs. max. pulsed collector current stipulated in the datasheet, but it comes with a note "Repetitive rate: Pulse width limited by max. junc temperature". I don't understand this. Is 25A a hard limit regardless of pulse width? I'm used to working with FETs that have much higher ratings (100's of amps).

I added a 15ohm resistor in series with the lamp, it still blew Q1 even tho the peak collector current would be ~13A, well below 25A. I've now added a 100ohm resistor instead and it's working reliably. The pulse on the gate looks clean ie. no horrible NGT on turn off. I'm going to try smaller series resistors and see what gives.
 

MisterBill2

Joined Jan 23, 2018
18,179
Re you able to try the system with the intended motor load? Or with a similar motor load? A motor is an entirely different kind of load from a light, so the results may be very different as well.
 

Thread Starter

Robin66

Joined Jan 5, 2016
275
Re you able to try the system with the intended motor load? Or with a similar motor load? A motor is an entirely different kind of load from a light, so the results may be very different as well.
Yes I'm sure the results will be different. But I want to understand the failure mode in a simple light-weight example first before I introduce a large inductive load with much larger peak currents.

It's been working reliably with 100ohm and 39ohm, but it fails after a few attempts with 15ohm (in series with the lamp so 25ohm total). This is quite frustrating. I'm definitely below the 25A abs max rating now, and I find it hard to believe that the junction is getting too hot within 10ms. And if that was the case it wouldn't necessarily blow when I first switch it on; it might blow after a few pulses.

I think this may be due to a delay in the driving circuitry being powered up after 240V is first applied. There's a reservoir cap on the 240AC-12VDC module which'll take a while to charge. In the time it takes the 12V line to ramp up perhaps Q1's C-B capacitance is pulling the base outside the 25V limit. Although surprised that 39ohm would be enough to prevent that. Zeners across Q1 B-E has already been suggested. Maybe a 100ohm resistor would work instead and prevent Q1 switching on at all during init.

I've now run out of IGBTs. I've blown 10 so far. I have more arriving tomorrow. Thank you RS
 

MisterBill2

Joined Jan 23, 2018
18,179
Yes I'm sure the results will be different. But I want to understand the failure mode in a simple light-weight example first before I introduce a large inductive load with much larger peak currents.

It's been working reliably with 100ohm and 39ohm, but it fails after a few attempts with 15ohm (in series with the lamp so 25ohm total). This is quite frustrating. I'm definitely below the 25A abs max rating now, and I find it hard to believe that the junction is getting too hot within 10ms. And if that was the case it wouldn't necessarily blow when I first switch it on; it might blow after a few pulses.

I think this may be due to a delay in the driving circuitry being powered up after 240V is first applied. There's a reservoir cap on the 240AC-12VDC module which'll take a while to charge. In the time it takes the 12V line to ramp up perhaps Q1's C-B capacitance is pulling the base outside the 25V limit. Although surprised that 39ohm would be enough to prevent that. Zeners across Q1 B-E has already been suggested. Maybe a 100ohm resistor would work instead and prevent Q1 switching on at all during init.

I've now run out of IGBTs. I've blown 10 so far. I have more arriving tomorrow. Thank you RS
You bring up an interesting possibility, which is that the IGBTs may be getting turned on but short of saturation. That would lead to a lot more power dissipated , probably far outside the safe operating area. So it may be time to rethink the control scheme, perhaps forcing a turned-off condition for a while until tghe driver is ready and fully charged. Since it seems that one area is not causing the failure, it may be someplace else. Situations like this is where a simulator could be useful, possibly.
 

geekoftheweek

Joined Oct 6, 2013
1,201
Rather than a slow comparator to protect the base from over voltage, how about shunt diodes, arranged so that they will go into conduction as the base to emitter voltage approaches the limit? Diode clamping can be a very effective means of protection. And it is simple to verify experimentally, as well. In addition, a shunt diode takes no extra power, it is functional even when everything else is switched off. BUT you do need to know the base voltage limit value.
I kind of just tossed that out after reading about the concept a few weeks back in a brushless motor controller project. I'm glad it got the ball rolling so to speak.
 

Thread Starter

Robin66

Joined Jan 5, 2016
275
You bring up an interesting possibility, which is that the IGBTs may be getting turned on but short of saturation. That would lead to a lot more power dissipated , probably far outside the safe operating area.
Yes that’s a possibility too. Maybe Vbase is not exceeding the 25V limit but it’s floating around a few volts.

I’ve been caught out by this kind of design flaw before. It’s not good enough that the circuit works in steady state. It needs to boot up first and survive that process
 

Thread Starter

Robin66

Joined Jan 5, 2016
275
Ok, nailed it. A 1k resistor between Q1's B-E didn't help. After desoldering another blown Q1, I set my floating probe to single capture and placed it across B-E. When I connected the mains it captured the first pulse as ramping up to 12V approx linearly in 10ms. It's just a coincidence that it takes 10ms for the 12V line to fully establish. I reprogrammed the PIC to delay 20ms after init before sending any pulses. This ensures that the first pulse is square. It now works reliably without any resistor in series with the lamp ie. 10ohm cold start up.

Got there in the end. Thanks for the help
 

MisterBill2

Joined Jan 23, 2018
18,179
Ok, nailed it. A 1k resistor between Q1's B-E didn't help. After desoldering another blown Q1, I set my floating probe to single capture and placed it across B-E. When I connected the mains it captured the first pulse as ramping up to 12V approx linearly in 10ms. It's just a coincidence that it takes 10ms for the 12V line to fully establish. I reprogrammed the PIC to delay 20ms after init before sending any pulses. This ensures that the first pulse is square. It now works reliably without any resistor in series with the lamp ie. 10ohm cold start up.

Got there in the end. Thanks for the help
When all of the obvious causes for a problem are found to not be the cause, it must be something else. Thanks for letting us know what it appears to have been. That indeed was a more subtle cause.
 
Top