IGBT Desaturation Detector

[TMaxElectronics]

A while back I posted about the gate driver I'm working on, and said I wanted to add desaturation detection to it. After reading about it a bit it seems that the most common method is the diode, current source and comparator method, and to a degree also differential voltage sensing and a comparator. But after thinking about it for a while I started wondering if a simple pnp transistor with a diode and two resistors wouldn't do the job just fine too. I was thinking of something like this (with a base resistor obviously lol):



And with one more transistor that could be improved to only activate the detection if the IGBT is actually supposed to be on:



I tested the circuit in LTSpice and there it seems to work out fine too (I added a 10us lowpass to the output):
 

So do you guys think this would work reliably in the real world? It seems to me like its quite a bit simpler to implement than the other methods.

[davekni]

The idea looks fine.  Might even be fine as drawn for hobby use.  Would need tweaks to be suitable for any commercial application.  For example, check device maximum specifications vs. simulation voltages and currents.  (At the moment I spot only one spec. being violated.)

[Mads Barnkob]

Desaturation detection has been discussed before, I think it was first brought up by Steve Ward on 4hv, but this is a VERY long time ago.

Back then, the conclusion seemed to be that desaturation detection was not fast enough to save an IGBT experiencing a huge overload in a DRSSTC.

Most of this is outside of my understanding, but I do remember the discussion about it

[petespaco]

Hello Mads and TMax.
  I, a novice at this, have been working on a  "DC to DC SSR Protector" myself,using an Arduino and a rail to rail opamp to amplify the voltage from a 75mv current shunt that I already use to drive an analog 50 amp current meter.
  MY isssue is that, with the ZVS Induction heaters, if I inadvertently push a workpiece too far into the work coil, I can easily create a momnetary excessive current spike.  I need a way to shut DC power OFF before the SSR fails.
(By the way, the IRFP260N's in the ZVS heaters hold up really well at 40 + amps continuous)

  It appears that the project that you guys are discussing relates to problems shutting OFF an IGBT (or Mosfet, I assume) in an AC  (resonant?) circuit.
 
Question 1:
If that is true, then, does this desaturation detector have to operate in one cycle?
Putting it another way, does the IGBT fail in only one overcurrent trailing edge event, or is the problem with eventual overheating of the die?

Question 2:
Why are aren't you concerned just as much about the leading edge?

I ask because, in my situation, I am using a DC to DC SSR to control power to my 2500 watt ZVS Induction heater.  All I want to do is to shut off power ASAP if I detect current flow in excess of what I set in software.

So far I have built and programmed the sensing device and bench tested it using a simulated voltage input.  I have options as to how much code I can really need, since I am sending info to an I2C LCD  (and to the serial monitor), which might even take  many milliseconds.
Once I do get things going, I won't really need the LCD or serial monitor code, but It's nice to see exactly what is going on.

[davekni]

Question 1:
If that is true, then, does this desaturation detector have to operate in one cycle?
Putting it another way, does the IGBT fail in only one overcurrent trailing edge event, or is the problem with eventual overheating of the die?

Depends on the device and use conditions.  In H-Bridge use, desaturation voltage can rise very quickly, making power dissipation extremely high.  The IGBT may last only a few microseconds at that power level.  On the other hand, with your ZVS induction heater, over-current may occur at the speed a person is inserting load into the work coil.  In that case, power dissipation may be only slightly over long-term capability, and millisecond speed is acceptable.  (ZVS oscillators can drop out of oscillation of load Q gets too low.  In that case, power dissipation may rise much more rapidly.)

Question 2:
Why are aren't you concerned just as much about the leading edge?

I think that circuit works fine at both edges.  It is detecting static desaturation, which can occur in the center of a gate-drive half-cycle when sine-wave current is at its peak.

[TMaxElectronics]

Regarding petespaco's post:
  The idea of this circuit is to protect IGBTs if the current rises beyond what the semiconductor is capable of conducting (regardless of what circuit it is in). In that case the voltage, and thus dissipated power of the IGBT will rise and the transistor could easily die.

  Basically the IGBT leaving the safe operating area :
 

  Depending on the type of overload the current could rise very sharply (for example a winding short in a motor) which as david said could kill the IGBT in microseconds. Because of that the only real option for me is to use a discrete circuit.
  All it does is to see if the voltage across the IGBT increases beyond a few volts when its on, and turn it off immediately if that happens.

  I'd say that for your application a few ms might be fine since you aren't trying to work against a possible short circuit. But if you want to make it as fast as you can, you can include a comparator and a DAC generating the voltage to compare the reading to.
  Then just use an interrupt to switch off the SSR.


I also improved my circuit a bit. Since I don't just want to use the driver in tesla coils or other resonant inverters I decided that I want to try guarding against output short circuits too. For that I added a bit of digital logic that immediately turns off the IGBT when desaturation is detected. In the spice sim that takes ~3us. A professor I talked to mentioned that delays of up to 10us would be acceptable for most bricks, which would make my circuit fast enough, but take that with a grain of salt since I haven't modeled the circuit very accurately.

The switch emulates a gate driver with an output enable input, and M2 is used to discharge the gate at a lower rate than normal to reduce voltage stress from the inductance during switch off.
I'll implement this in another test version of my gate driver soon... once my exams are over

[Twospoons]

That SOA curve looks a bit wrong. Usually they're shown in datasheets like this:


As you can see the SOA is also a factor of time.

[T3sl4co1l]

I strongly endorse the use of such a circuit.  I've been using them since my earliest work in power electronics (above 1kW), and would still be able to hold all the devices I've destroyed, in two hands.  That includes one big industrial module, which is the main reason they're probably not all in just the one hand.

I always used a comparator circuit, which provides more accurate measurement of Vce; this might be of interest for crude current monitoring (because Vce depends on Ic) or for use with MOSFETs (which benefit from exactly the same treatment, they just don't need to respond so fast -- thanks to the lower power density, 20-100us may be adequate time).  For detecting the rapid rise on leaving saturation, nothing fancy is needed at all, and indeed a few BJTs will do the job.

A note: you'll want a B-E resistor on the "sense" PNP, to prevent it from accidentally turning on.  Also mind the capacitance and reverse recovery of the sense diode: there's massive V and dV/dt on the other side of it.  The series base resistor helps, but I would consider a E-B clamp diode, and maybe RC filter as well.  Also check the relative values of collector and base resistors; I assume these values are just for a quick simulation, not an actual build, so just remember to check this later.  (The high collector current, ~50mA, is rather hungry for a gate driver; but that can come down substantially, no problem.)

Note that any filtering you add, slows the response.  You can afford a few microseconds, as long as it's still a modest fraction of the total switching waveform.  Too much, and obviously, there's just nothing to sense, you'll need a different method.

Back then, the conclusion seemed to be that desaturation detection was not fast enough to save an IGBT experiencing a huge overload in a DRSSTC.

It is a perfect counter to the situation it targets.  There are other situations that result in failure, however.  For example, plain old overheating -- whether from switching too slowly (large t_r, t_f), or too often (high Fsw), or drawing greater than rated load current (but still without going into desaturation).

To cover these, the basic method is to simply limit peak, average or other load current.  Monitor with a suitable sensor (current transformer, say; Hall effect sensors are pretty slow, but also adequate when Fsw is low) and shut it down if it exceeds nominal.

For more finesse, you'd unfortunately need a more complex method, perhaps a real-time measurement of device voltage and current, the product of which is fed into an equivalent thermal impedance model, to infer real junction temperature.  This is a PITA because analog multiplication is hard.

That SOA curve looks a bit wrong. Usually they're shown in datasheets like this:

As you can see the SOA is also a factor of time.

The highlighted region corresponds to the leftmost column; I suppose arguably, as long as it's under the DC curve, one could say time actually isn't a factor.  But that changes quickly at only little higher voltages!

Tim

[TMaxElectronics]

For more finesse, you'd unfortunately need a more complex method, perhaps a real-time measurement of device voltage and current, the product of which is fed into an equivalent thermal impedance model, to infer real junction temperature.  This is a PITA because analog multiplication is hard.

That's actually kind of what I was going for eventually
There is some good info about pcb based rogowski coils out there, which I should be able to make fit around the terminals of IGBT bricks. AD also sells some fully integrated analog multipliers with bandwidths of up to 2G (), which should be good to get the momentary power dissipation in the die. Then all that would be needed is a leaky integrator (with the leakage speed being the equivalent of the thermal impedance, the target voltage being the heatsink temperature) and you'd have the thermal energy currently in the die... at least in theory
Issue with that is of course that it would require quite a bit of tuning for each IGBT and I kind of want to make a universal driver. Maybe I'll just put some expansion headers on the driver and make one board with just the desat detection, and one with the full thermal model to plug into it

The measured energy could then be put into a comparator to decide if an emergency switch off is necessary or not. Though I suspect it is a better Idea to just set an error flag and wait for external switch off since hard switch off would cause a lot of additional thermal stress I'd assume. It might even be cool to put one of SI's fully isolated ADCs into the thing and a micro on the low voltage side that would allow for readout of the data via RS485 or something like that. I do have an 8 bit PIC there as the flyback regulator, I could just use a slightly larger one and add that functionality to it. That would even allow for dynamic tuning of gate drive voltages and stuff.

For now that will all have to wait though since I have some exams to write

I assume these values are just for a quick simulation, not an actual build, so just remember to check this later.

Yeah I really just wanted to know if the idea I had in my head actually worked or not The final design will need quite a few additional/different value components.

[T3sl4co1l]

Can certainly be done analog, but you're probably as well off using a 20MSps+ ADC, and a nice MCU, or FPGA, so that all the parameters can be set at once.  Bonus, you can add a readout function to monitor live gate/collector waveforms as a crude oscilloscope.

Hard switchoff is really only important when already under fault conditions.  Under normal operation, adding another couple mJ on top of an already hot junction, makes very little difference.  It's when the current is so high that normal turn-off generates a huge overvoltage (even just due to lead inductance), that you may need a softer driver.  (A number of drivers provide this functionality, with a separate "fault turnoff" output that you can put a larger gate resistor on.)

Tim

[TMaxElectronics]

but you're probably as well off using a 20MSps+ ADC, and a nice MCU, or FPGA

Yeah I was thinking about that too, but that would make the circuit a lot larger I think. The FPGA would be at the very least 48 pin, with any micro that is fast enough being similarly large or larger. Then again the analog might need a negative supply (though the flyback could do that without issue I guess). The analog approach would only need like 3 opamps and the multiplier (which is a single SO8). And I could probably just make a little spreadsheet to turn the IGBTs parameters from the datasheet into the required component values.

Plus a friend of mine works for AD so I can get small amounts of their chips for free

(A number of drivers provide this functionality, with a separate "fault turnoff" output that you can put a larger gate resistor on.)

Yeah I saw that on the drivers I got from Steve together with the huge IGBTs. I think I'll add that to my driver, with a kill input that latches an error state until the next rising edge of the control signal.