IGBT Gate dv/dt

Eugeneo · Joined: 30 Aug 2005 Posts: 155 Location: Calgary, AB

A little off topic but there seems to be an abundance of bright people here so I just thought someone could give me some hints.

I have a charge system that charges 4 x 350v 22uf caps and then discharges them to 1kv aircraft igniter tip (This app is not for an aircraft but it's designed to work under water).

Operation:
Once the voltage is at 1kv, I have a Pic triggering a optoisolator. The optical isolator has a max rise and fall time of 20 us (detector side).

That output is triggering the gate of a completely isolated IGBT at 15 volts. The IGBT rise is 60 ns and fall of 230 ns.

Problem:
The system functions the way it is supposed to. Except once in a while during my tests, the IGBT will completely latch up, and must be replaced.

I've checked the safe operating zone for the IGBT and everything is in check. The temperature of the IGBT is only warm after several minutes of operation. The IGBT is good for 70 A continuous at 1200V, 230A pulsed, short circuit of 10us.

My thoughts:
Is it possible that my optoisolator is triggering too slow and causing a high inrush of current when the collector/emitter has not completely reached a turn-on state, causing excesive power dissapation on the IGBT?

The igniter tip kind of acts like a metal oxide varister and triac at the same time. when the voltage applied to the tip is 1000volts, it will discharge the caps between 5 to 10 us. If the voltage is lower than 400, the tip discharges the caps within 3 seconds.

Any and I mean any suggestions would be appreciated. Thanks in advance.

kender · Joined: 09 Aug 2004 Posts: 768 Location: Silicon Valley

Eugeneo · Joined: 30 Aug 2005 Posts: 155 Location: Calgary, AB

libor · Joined: 14 Dec 2004 Posts: 288 Location: Hungary

Brian S · Joined: 06 Sep 2005 Posts: 13

I just went through the design loop for a 500V 1A high-speed high-side FET switch....similar application. I agree that using an Opto Isolator won't work. Gate drive voltage and current are critical. But using a Gate Driver chip directly leaves you with the problem of supplying power to the driver chip, as it must be isolated.

My solution (untested as yet!) is to power the Gate Driver from a 12-15V supply referenced to the PIC's power/Gnd, and drive the IGBT/FET with a gate drive transformer. Check out the Triad GDE25-1 (Digikey) or similar. A driver in the TC4426 series may work, but larger drivers are available.
I needed PWM of the FET, which presented another problem. IR has a valuable App Note (AN 950) on gate drive circuits for PWM to avoid the issue of gate drive sag at low duty cycles due to insufficient inductance in these high frequency transformers. Cost is low, and it assures stable drive levels under any (reasonable) conditions.

As to the crazing of your glass container: you might be able to find a heavy-wall polycarbonate beaker or tub. The polycabonate is a good insulator and is mechanically very tough - much more shatter resistant than glass. Try not to drill or cut it; machining tends to create stresses that can lead to later cracking near the cut.

Good luck!

Eugeneo · Joined: 30 Aug 2005 Posts: 155 Location: Calgary, AB

Eugeneo · Joined: 30 Aug 2005 Posts: 155 Location: Calgary, AB

Brian S · Joined: 06 Sep 2005 Posts: 13

Eugeneo,

For normal PWM frequencies of <20KHz, the PWM module of the PIC works well; You can specify PWM duty cycle directly. Or the PIC can output a single pulse by using the "delay" function. My application is rather specialized, running at 20-50MHz; I'm using a dual DDS chip with some gating to run at the higher speeds. Though I'm not using a driver chip at these speeds, I have used the drivers at 20KHz in the past with good results.

The transformers designed for this service are inherently wideband. The Rise time is therefore a function of the gate driver's output characteristics. Compared to your Opto drive, you'll be going from >5mS(?) gate switching to <50nS switching.

These transformers have quite low leakage inductance, so there doesn't seem to be a problem with inductive effects; a small signal diode could be used at it's primary to clamp any return spikes toward the driver.

The inductive surge in my app is taken care of by the intrinsic diode of the FET. If you rely on that, be sure to pick a device that's Avalanche Rated, and pay attention to peak pulse current.

Many IGBT's have a similar structure composed of the IGBT and a HEXFRED in the same package. Or you can use an external Power Schottky to help out. Either way, be sure to run the numbers for dissipation; the diode will produce significant heat, often more than a low Ron FET (but probably less than an IGBT).

Don't forget to wear your safety glasses!