"For those of you who are going to tell me the gate of the IGBT needs to be grounded - it's working without the ground. " - Then there is definately a problem with the IGBT... think about it, how is the gate capacitor discharging ? Sounds like the IGBT G-E junction has been compromised.

 

Also why are you designing with a component that is being phased out and is obsolete?
https://www.digikey.com/product-det...ogies/IRG4PSH71KDPBF/IRG4PSH71KDPBF-ND/811687

Anyway ... looking at the specs of the IGBT, the input capacitance is 6.8nF typical .... with your original circuit with the NPN PNP drive transistors that creates a "ringing" current spike across the gate capacitor of 2.56 Amps which is probably what damaged the IGBT ... placing a 100 Ohm resistor between the NPN PNP emitter output and the Gate of the IGBT lowers the current across the cap to 200mA and reduces ringing considerably.

 

@Beau Schwabe Is there a way to test the IGBT? Like with an ohm meter, diode check or something?

 

Put your DMM on diode check and measure the "Gate to E" and the "Gate to C" then swap the leads to the DMM and do the same test. You can also check "C to E" swapping the leads as well, but be aware that it can turn "ON" from the DMM check on the Gate (voltage from the meter). However if you ground the GATE to E, then the check from C to E it should be "OFF".... ALL of this has to be done when the IGBT is not connected to anything else. You can always grab one that is "known to be good" and see what to expect.

Note: On the test above, the "Gate to E" and "Gate to C" should indicate that the diode is OPEN ... usually a reading greater than 3.0V ... anything less and you have a short through the Gate.

 

DMM on diode check:

-G > +C = 0.643v
-G > +E = 0.028v

+G > -C = 0.406v
+G > -E = 0.028v

-C > +E = 0.391v
-E > +C = 0.635v

Does this look like the gate and emitter are somewhat shorted?

Resistance gate to emitter: Meter wouldn't stabilize. Kept jumping around between about 400Ω and zero ohms.

Have another IGBT on the work bench and it appears to be open in nearly all cases. Only -C > +E showed 5 MEGΩ.

Maybe that's why the 555 couldn't trigger the gate on the IGBT. I'm removing the FET and going to try striking the gate with the output from the 555 through a 100Ω resistor. Let me know if you think that's wrong.

 

"Have another IGBT on the work bench and it appears to be open in nearly all cases. Only -C > +E showed 5 MEGΩ. " - Those are the readings I would expect.

 

Well, after having removed the FET and connecting the 1KΩ to the IGBT Gate - it works as expected.

Thank you so very much for helping solve this problem. Going to leave it running at 50% for a while and see what kind of temperatures I get. Originally when the 555 output was tied to the gate via the 1KΩ the 555 would hang up around 10 volts and wouldn't do anything else. I assumed (and we know what happens when we assume) the IGBT took more power than the 555 could deliver. It's been like that from day one. I assumed the IGBT was good. Never assume!

Thanks again for the help.

With the 1K the square wave is flat across the top and slightly like the tailing off of a capacitor on the bottom side coming out of the 555. When tagged to the gate of the IGBT it looks like shark fins (triangular waves). I guess that's the capacitance in the IGBT. Nevertheless, I get full sweep from near 0% to near 100%. Already ran a light bulb. Now running a small 12 V fan. No sign of heating by the IGBT at all. 77˚F after a few min's. at 50% operation. Will try near zero and near full just to see. But by this point things were heating up.

Will post the final diagram shortly.

 

100 Ohms not 1k

If it were a MOSFET, the "shark fins" would cause the MOSFET to spend more time in its linear mode resulting in HEAT.
Since this is an IGBT the transistor output turns ON early in the "shark fin" and OFF late in the "shark fin" so it may be OK. I would still rather use a 100 Ohm as opposed to 1k on the gate however. It will lessen the "shark fin" to more of a square wave.

 

Final diagram - working beautifully. No ringing, no overheating.

 

Will change that up.

 

I'd be interested to see the scope of the voltage across the 1K (or 100ohm if you change it).

 

Wish I took a picture with the 1K. Have changed to the 100Ω as suggested. The square wave is more squarish with the 100 when looking at both the gate and the output (555). With the 1K the wave was much sharper on pin 3 (555 output) and like a shark fin wave shape on the gate. Going with the 100Ω has evened things up, but the wave doesn't look all that impressive. Maybe that's not important.

 

1 / ((10 µS x 5 div's) ÷ 1,000,000) = 20,000

20KHz. No???

 

Why do I get a funny feeling you have a near short to ground in parallel with your IGBT gate?

Perhaps because it would explain everything you have been saying.

Bob

 

@BobTPH I don't know why you get that feeling. I did find that the IGBT I was using was bad. Drawing a lot of current even when it wasn't switching ANY load. I've had it running at 50% power 50 minutes now and the IGBT, running an automotive lamp has heated up to 93.5˚F Warm but understandable. The lamp draws a bit of current. It's the brake light portion of an automotive bulb. The bulb housing is at 94.5˚F, only one degree hotter than the IGBT. When I get around to running a blower motor I hope to see little if any heating beyond what it is already carrying for wattage. Ambient room temperature is 76.5˚F. So that's 18˚ change. I'm not excited about that but it's a lot better than the original problem where I was burning up 2N3906's. Actually only one burned up. But a FET got pretty hot too. Somewhere around 122 degrees.

 

Running the lamp at full power has allowed the IGBT to cool a bit. Down to 87 degrees, and cooling.

 

I don't know why you get that feeling. I did find that the IGBT I was using was bad.

Yep, that is what I was getting at. The gate that sucker was blown, shorted to the emitter in likelihood.

I posted my conjecture before I read that you had gotten it working.

I would take out even the 100 Ohm resistor to get sharper on and off and less heating. Maybe 10 Ohms. Insulated gates are not like the base of a BJT needing a current limiting resistor. The 10 Ohms people use sometimes is to minimize ringing.

Bob

 

Thanks @BobTPH. I DID have a 1K there before. It had a very sweet square wave on pin 3 of the 555. However, the waveform on the IGBT Gate looked like a shark fin. Having gone down to 100Ω the waveform looks like that post #72 above. Not seeing any ringing either.

What about my math? The scope is set to 10 µS sweep and there's 5 major divisions per period. Did I come up with the right number? 20KHz?

 

I DID have a 1K there before. It had a very sweet square wave on pin 3 of the 555. However, the waveform on the IGBT Gate looked like a shark fin.

The current into the gate is proportional to the difference between these two waveforms. I would expect them to be almost identical except right near the transitions.

What about my math? The scope is set to 10 µS sweep and there's 5 major divisions per period. Did I come up with the right number? 20KHz?

This sounds right to me!

 

A resistor connected to a gate is an RC low pass filter.

Bob