I'm a Tesla enthusiast and build Tesla coils now and then. One thing I have always wondered, a lot of designs (some very popular) utilize a transformer to drive a N-MOS half bridge (see image below). My concern with this, and simulations have shown it to be true, shouldn't this type of a design result in a high amount of shoot though since there is no dead time incorporated?

With that being said, I would love to get away from bootstrapping my driver and forcing it to be sitting on the high voltage rails (debugging and inspection with a scope gets much easier). My problem is I do not see an easy way to incorporate dead time.

The only method I see to add dead time and have galvanic gate drives would be to use a full bridge driver to sink and source current to two independent drive transformers and hackishly bypassing the bootstrap drive tying it to vcc.

So my questions are:

  1. Do you agree with my assessment that for switching at 200khz, the design illustrated should have the bulk of its power losses due to shoot though?
  2. Does anyone see a way with to easily employ dead time with galvanic isolation without using a full bridge driver as explained above?

enter image description here

  • \$\begingroup\$ P.S. My question has a bad title in my opinion... Feel free to make it better if anyone can better describe what I am asking. \$\endgroup\$ – MadHatter Oct 5 '16 at 1:42
  1. No. The two 1n4148 diodes ensure that when the gate drive transformer secondaries start to reverse their polarities, the currently conducting IGBT turns off faster than the currently blocking IGBT, creating a very bried period where both are off (the dead time). This is because the parasitic gate capacitance of the IGBT to be turned on has to be charged trough the 6.8 ohm resistor, while the gate charge of the opposite IGBT can bypass the resistor by flowing trough the diode.

  2. Fully isolated MOSFET/IGBT drivers for this purpose do exist, if you need more control over the transistors and/or if you want even faster switching. Digikey lists over a thousand isolated gate drivers, e.g. ucc21520

  • \$\begingroup\$ I guess that shoot though is more of a spectrum rather then discreet problem. While the shoot though is better with dead time, the diodes do make a huge difference vs only resistors... I'll look into isolated drivers, otherwise I may just do a simpler design as shown. All my previous drivers have been bootstrap drivers referenced to the negative power rail. \$\endgroup\$ – MadHatter Oct 5 '16 at 19:32
  • \$\begingroup\$ @MadHatter I don't understand what you mean about shoot trough being a spectrum. With sufficient dead time, you can completely eliminate shoot trough. Too much dead time will however cause its own problems which also cause losses (heat) in the transistors (by interfering with zero voltage switching and due to the reverse recovery time of the diodes within the IGBTs). If you want to understand more, search for application nodes about "LLC resonant converters". They are series resonant half bridge DC-DC converters which operate extremely similarly to a double resonant half bridge SSTC. \$\endgroup\$ – jms Oct 5 '16 at 19:54
  • \$\begingroup\$ I mean that the diode does not completely remove shoot through, I guess you could tune it out by selecting the resistor to be big enough to completely turn off the low side before the high side reaches conduction threshold... But then your rise time suffers and you end up with higher switching losses anyways. \$\endgroup\$ – MadHatter Oct 5 '16 at 19:58

Various isolated gate drivers. However keep in mind that you will need at least one isolated (highly isolated) PSU to drive the gates. Also the switching frequency that is acheived using pulse transormers is quite high compared to opto-coupled gate drivers. IMO, you should use some driver with magnetic coupling or some other kind that allows high frequency, opto isolated aren't suitable for your application.

Better and cheaper, would be to have two gate driver circuits and separated pulse transformers. Then you have to add an interlock circuit between them.

Probably there are some other SSTC circuits, more complex, with a cross conduction delay, so I think you better stay with this basic circuit as is.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.