I'm designing a capacitive discharge circuit. I'd like to get some feedback before I buy parts for testing. The output should be around 100v and thought an IGBT would work great in discharging the capacitor. But I'm unsure. What I'm hoping to do is switch the 100v on with a low voltage to the gate from the optocoupler. How much voltage would I need to the gate? I have a couple of question marks on my design below indicating problem areas that I'm unsure about. Please advise.

I'm open to suggestions and criticism. revision 6 of cd circuit

  • \$\begingroup\$ I am not sure the discharge BJT will work at all. It seems your step-up transformer is insulated, so there is no clue whether the power +12V is actually at a higher voltage than HV RTN. I can't help with the IGBT though. \$\endgroup\$ Commented Jul 19, 2014 at 7:55
  • \$\begingroup\$ Moreover that TIP47 is voiding the insulation. You should use a relais or another optocoupler. \$\endgroup\$ Commented Jul 19, 2014 at 8:01
  • \$\begingroup\$ What will (or should) the voltage on the IGBT gate be when the PC713 is not conducting? \$\endgroup\$ Commented Jul 19, 2014 at 11:44
  • \$\begingroup\$ Voltage at gate should be zero when PC173 is not conducting. \$\endgroup\$ Commented Jul 19, 2014 at 14:42
  • \$\begingroup\$ Then make it zero! As it is now it is floating, so the IGBT could start conducting at every full moon (or any other inconvenient moment). \$\endgroup\$ Commented Jul 19, 2014 at 15:38

1 Answer 1


For the IGBT driver, I suggest an isolated supply (galvanically isolated DC-DC converter) and opto-gate driver. One of the easiest ways to drive the gate is with an isolated gate driver such as the Avago ACPL-312U-000E.

You need to bring the IGBT gate positive by something like 15V relative to the emitter to fully turn it on, and to drive it to zero (or a few volts negative) to fully turn it off. That should be done fast if you want the transistor to turn on snappily. It should never exceed something like +/-20V relative to the source. What you have there will not achieve that as the 15V from your voltage divider is relative to your load - terminal and not the IGBT source, so the gate will likely be damaged.

It's possible a variation of your circuit with the HV section flipped (load on the IGBT collector to +100V) could be made to work.. but the IGBT gate has no path to discharge the charge so it will leak off slowly, which might be disastrous.

If this is a welder you're making, keep in mind the load will short the 100V supply (the 1N4005 won't help) when the IGBT is on.

Edit: you can derive an appropriate supply voltage from the 100V for the IGBT driver as follows:

It supply between 0mA and 5mA without allowing the gate voltage to exceed the absolute maximum 20V voltage.


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ @Stephro The dc-dc is isolated. See <a href="emcohighvoltage.com/pdfs/aseries.pdf" title="A01 Datasheet">datasheet</a>. The use of the IGBT gate driver is probably the way to go. I think I understand the IGBT a bit better now. Putting the load on the collector makes sense. But you say the gate no path to discharge? Darn these switches aren't perfect. How can I fix this? Drop down resistor to ground? \$\endgroup\$ Commented Jul 19, 2014 at 16:13
  • \$\begingroup\$ The resistor would work, but it would be slow. The gate drivers have push-pull outputs so they'll pull the gate charge out quickly. \$\endgroup\$ Commented Jul 19, 2014 at 16:16
  • \$\begingroup\$ Updated schematic to include the IGBT Gate Driver. Any feedback would be helpful. \$\endgroup\$ Commented Jul 19, 2014 at 20:42
  • \$\begingroup\$ Vcc needs a source of about 15V @ 5mA -10mA relative to emitter. Vee should connect to the emitter. Voltage divider won't give you enough current, and it's connected wrong. \$\endgroup\$ Commented Jul 19, 2014 at 22:11
  • \$\begingroup\$ Spehro I looked at data sheet and see that Vee should connect to emitter. But can you elaborate on connected wrong? Your guidance would be helpful... \$\endgroup\$ Commented Jul 19, 2014 at 22:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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