Quick question about dead-times

[Zilk]

Hello,

i was just curious about dead-times with a GDT.
In some schematics there is a parallel diode across the gate drive resistor and in some don't.
For example in Mads plan for his DRSSTC I there isn't one.



In Steves DRSSTC 5 there is.



As far as I understand, this diode should shorten the switch-off time to introduce a dead-time, am I right?
How can know if i need one or not?

Would be thankful, if someone could explain this to me.

Greetings

Kai

[ritaismyconscience]

You should use the diode. It turns the IGBT off faster and makes sure that the IGBT is off when the other IGBT turns on.

[Weston]

Having both the top and bottom switch turn on at the same time would lead to shoot through as the two switches try to conduct current at the same time, effectively shorting out the supply. This can lead to excessive power dissipation, voltage spikes, and destruction of your IGBTs. Deadtime is typically desirable to prevent this from happening.

With the gate drive transformer it is difficult to add dead time directly to the drive signals, so the diode is added to effectively remove the diode and lead to the gate turning off faster than on. Basically adding dead time.

If you look at a IGBT datasheet ( for example https://media.digikey.com/pdf/Data%20Sheets/Fairchild%20PDFs/HGT1N40N60A4D.pdf) you will see that the turn on delay is less than the turn off delay, which would seem to imply that the diode is needed to prevent both switches from turning on a the same time. However, these switching times and delays are measured without negative gate bias, the gate drive waveform starts at zero and the turn on threshold is only a few volts, so there is a larger voltage difference between the on voltage and the threshold voltage than the off voltage and the threshold voltage. With a GDT the voltage applied is symmetrical and goes negative, so the turn on delay and turn off delay is going to be more closely matched, with turn off delay most likely being less than turn on delay, even without the diode, hence why some designs do not have the diode.



You could calculate the turn on and turn off delays based on the gate resistance, gate capacitance/charge, and gate drive voltages. To a first order approximation the gate is just a capacitor. However, either way, with or without the diode, it would probably work. If you are concerned about the gate drive, it would probably be worth scoping out, but that would require an isolated oscilloscope probe for the high side gate.

DRSSTCs get close to zero current switching so switching losses are small relative to the conduction losses. You can get away with a lot on the gate drive unless it looks really horible.

[davekni]

Even with gate-drive that's symmetric around the threshold voltage, IGBTs have slower turn-off due to base storage time.  IGBTs are mostly minority-carrier devices, creating turn-off delay as the carriers recombine.  FETs don't have that extra turn-off delay.  Generally, newer IGBTs have lower base storage time than old ones.  Even with the latest process technologies, there's a trade-off between fast base turn-off and low Vce.

[T3sl4co1l]

The transformer also has a lowpass characteristic, limiting rise/fall time in general.  (This can be improved by using a twisted-pair transmission line transformer, with more pairs in parallel to lower the characteristic impedance between the windings; and by using a thicker core, to use less winding wire length.  Thinner wire can also be used, as the RMS current is not very high; though the isolation voltage still is.)

Tim

[Mads Barnkob]

I am not quite sure if this application note is directed only towards high-side drivers or also does well for GDT, it only mentions that core-less transformers are preferred.: https://www.infineon.com/dgdl/Infineon-Deadtime_calculation_for_IGBT_modules-ApplicationNotes-v01_00-EN.pdf?fileId=db3a30431a5c32f2011a5daefc41005b

It has a walk-through of adding the different components in the gate circuit.


This old thread might also be of interest.: "GDT Experiments" https://highvoltageforum.net/index.php?topic=588.0