Quote:
Originally Posted by energymiser
Hay there. As I am slowly wrapping my head around this project, questions are starting to come to mind. There has been discussion about shoot through and ways of stopping it. This seems to be a potentially expensive problem affecting the longevity of the IGBT's. From what I can tell the software dead time between high and low side switching is just part of the solution.
There was some discussion about the gate resistor values and their effect on something called "autocommutation" or shoot through by a sort of back door. Obviously something I don't really understand but was wondering if there was some sort of consensus as to the values of these gate resistors?
Have a great day.
|
Well... you asked... so pardon the book... Hopefully I get all the details right, or at least close enough to illustrate the point.
Yes, shoot through is a problem. Just to be clear: shoot through is when both the high side and low side IGBT are both "ON" at the same time. This shorts the + and - bus bars together. It's called shoot through because the current runs practically unhindered between the + and - side. This really heats things up and wastes power. And so people go to great lengths to prevent it from happening. One way to help solve it is to make the microprocessor incapable of turning both sides on at once. The dsPIC chips have such a mode. You can set it so that you cannot tell it to turn both the high and low side on at once. Another way is to create some dead time where theoretically neither IGBT is on. This allows the currently "ON" side to properly turn off before the other side turns on. You don't want to make the dead time any longer than absolutely necessary. The reason for this is that when no IGBTs are on the motor will still try to inductively kick back as the field collapses. If no IGBTs are on then the freewheel diodes must absorb this energy. Say hello to heating problems again. So, the dead time should be long enough but no longer.
You probably are wondering, then, "how long is long enough?" Well, that's where gate resistors come in. Slight side track: IGBTs are turned on by applying voltage to their gate. The gate has a very high impedance (millions of ohms) but it still has capacitance. You have to charge the capacitance in order to turn the gate on. Most IGBT documents will list the gate charge in coulombs (well...nC but still...) Let's say a particular IGBT has a gate charge of 3000nC. You need to charge that to turn on the gate. If you use +15V and a 5 ohm resistor then you can provide a peak amperage of 3A to the gate. 3000nC / 3A = 1uS charge time. The converse is also true: You need to remove all that charge to properly turn it off (well... it will start turning off before you remove it all but you still are going to being removing it all). If you do nothing to remove the charge then it will just bleed through the gate resistance. Since the gate is very high impedance this will take forever (in IGBT terms... forever being mS probably) so you want to drive the gate with negative voltage to turn it off fast. This next part I'm not totally sure of so someone correct me if I'm wrong but: If you use a negative voltage then you technically are now way under the charged gate voltage. If the gate was charged to 15v and you have -7.5v then the difference is 22.5v. 22.5/5 is ohms is 4.5A. This will make the gate drain in about 0.66uS. so you could probably use a dead time of about 1uS and be safe. In the firmware for the controller I used a dead time of 2uS. That might be a bit too much but we'll see.
So, the next logical thing to wonder is: why not just use like a .5 ohm resistor and turn it on and off really fast? Easy turbo! You could damage things by trying to slam the gate on too fast. The IGBT document will likely have the minimum gate resistor they want you to use. Don't go under that. Big IGBTs can use somewhere around 1 ohm resistors and smaller ones 3-6 ohms.
If your resistor value is too small for the IGBT you are using then it is possible that your calculated turn off time from above will be shorter than the reverse recovery time of the IGBT. If this happens then the IGBT will still be conducting for a short time while the other IGBT is on. That's bad. So this means you once again do not want to use a resistor value smaller than what the documentation told you to use.
And, the auto commutation thing: The gate and drain of the IGBT have parasitic capacitance and this can cause the gate to spontaneously turn on when big voltage swings happen on the drain. However, if you are currently driving the gate off with negative voltage then you almost certainly won't have this problem because you are actively removing the gate charge. It's only really a big problem if you leave the gate floating when not driving it on. So, don't do that. The design of our 3 phase controller includes negative gate drive so it shouldn't have a problem with auto commutation.