In paralleled IGBTS, what is the best way to monitor IGBTs?

[kappa_am]

Hi all,
When IGBT's are parallel we cannot discriminate failed IGBT measuring GS and DS voltages; I was wondering what is the best way to discriminate failed IGBT in a parallel set?

Thank you.

 

[KlausST]

Hi

measure each IGBT's current via a small series resistor.

Klaus

 

[kappa_am]

Thanks for your response. is there any other way? I don't want to add resistors to the circuit. it's about 2400A circuit, even a small resistor generates lots of losses, besides extra cost.

 

[KlausST]

Hi,

Hall sensor, current transformer (a simple one).

 

[FvM]

I didn't yet see silently failed high power IGBT. Did you?

Consider that industry standard IGBT modules have multiple chips in parallel. There's neither a feature to check individual transistor chips.

 

[schmitt trigger]

That is what I was thinking....
I've never used a 2400 amp device, but have used 200 amp ones.

When they fail, they fail spectacularly.

 

[c_mitra]

... I don't want to add resistors to the circuit. it's about 2400A circuit, even a small resistor generates lots of losses, besides extra cost.

But the series resistor need not be large- it will get you another added bonus- a small series resistor will have a tendency to equalize the current. Without that, you will never be sure that the current is fairly shared by the parallel transistors.

AND INEQUALITY in current sharing may be the one of the main causes of transistor failures?

 

[FvM]

Is it a real design problem or a theoretical consideration? In the former case, are you paralleling industry standard IGBT modules or a large bunch of medium power IGBTs?

All IGBTs have a positive Vce,on TC at higher currents which allows paralleling without additional balancing means. For optimal current sharing, this design suggestions by Infineon should be considered:

symmetrical design of IGBT current paths (identical stray inductances)

symmetrical design of gate driver (same driver stage, seperate gate resistors,

splitted Rg with appr. 1/3 of it in the emitter leads)

use devices of one production lot (smallest parameter deviations guaranteed)

symmetrical cooling conditions (identical heat-sink temperature and flow rate
below the paralleled devices)

These conditions are already met in a IGBT module made of multiple IGBT chips.

In the lab, you can determine equal current sharing by measuring the voltage drop between the common emitter rail and individual emitter sense pins. Or dynamically with a small Rogowski coil or an open path dI/dt current sensor.

 

[kappa_am]

Thank you all for sharing your opinion. I always get great help from the forum.
actually, The total current is about 1800A, I am designing for 2400A using 480A switches. I want recognition method for adding fault toleration capability to the inverter. Though, I have my earplugs.

- - - Updated - - -

But the series resistor need not be large- it will get you another added bonus- a small series resistor will have a tendency to equalize the current. Without that, you will never be sure that the current is fairly shared by the parallel transistors.

AND INEQUALITY in current sharing may be the one of the main causes of transistor failures?

The kind of switch I am using has a positive thermal coefficient, so there is no need for series resistor and current sharing would be adjusted automatically.

So, the only way is to roughly measure the currents? isn't there any method to use phase current instead current of each IGBT? something like wavelet analysis ,... what will be the effect of failing one IGBT on the shape and harmonics of the current?

 

[FvM]

I keep my point that you don't have a realistic idea how an IGBT usually fails. Perhaps for cases where the IGBT not actually fails, but a gate driver stops switching. Or a loose collector bus-bar that luckily disconnects completely instead of catching fire. Very rare.