DRSSTC low power testing killing IGBTs

[FonziDaytona]

I see the on board MMC has bleeder resistors and my upgraded MMC doesn’t. Maybe startup IS the issue after all?

-Matt

[davekni]

I see the on board MMC has bleeder resistors and my upgraded MMC doesn’t. Maybe startup IS the issue after all?

That is also the best guess I'd have.  Increased GDT turn count will make slower ring when startup fails.  That might contribute to failures.  Though I haven't figured out exactly how startup failure would result in fried IGBTs.

[Mads Barnkob]

I see the on board MMC has bleeder resistors and my upgraded MMC doesn’t. Maybe startup IS the issue after all?

-Matt

This is what is tried to be fixed with a 10K resistor across the inverter output, to get rid of any residual charge that could confuse the logic at start of a pulse, not just start-up, but it would sound sputtering and unstable.

Did you take a look at all gate driver resistors, TVS and diodes from GDT to IGBTs? Many components can fail "half" at IGBT failures, like gate resistors going from 10R to 100R-10K and you get all these weird behaviours from differently damaged components in the drive chain.

[FonziDaytona]

Here’s a picture of the setup. It’s a pretty standard design…

As for TVS testing, I’ve tested for short circuit and none are shorted. To test open, I’d need to remove and connect to my hv power supply, right? Any other way to guess?

Zener diodes all test 0.6v, so guess they’re good. Gate resistors are 5.5ohms, so those should be good.

I’m itching to tweak something (easy that is) while I wait for replacement IGBTs :-) Should I remove some turns from the GDT? Or, should the bridge resistor negate any shenanigans from it having higher inductance?

Another thought that I had was voltage sag caused by smallish bus capacitors (1000uF each) and a much higher MMC (.165uF vs .075uF) than designed?

-Matt

[davekni]

It’s a pretty standard design…

Probably unrelated to failures, but I see IGBT Kelvin emitter connection is defeated by shorting the two emitter terminals together on the ECB.  That increases internal IGBT die Vge spikes, especially given hard off-to-on switching of this driver that doesn't support phase lead.  Perhaps the relatively-low normal Vge of this design prevents internal Vge spikes from becoming problematic.

To test open, I’d need to remove and connect to my hv power supply, right? Any other way to guess?

Vce TVS diodes can be tested with IGBTs removed.  Apply some voltage to half-bridge supply, say 40V.  Then verify that output (Out+) gets clamped around +440V and -400V.

I’m itching to tweak something (easy that is) while I wait for replacement IGBTs :-) Should I remove some turns from the GDT?

Your decision.  I'd remove turns at least to make it like the original design with 10 turns.  If it were my coil, I'd change to two twisted pairs as well.  But there may be some comfort in exactly matching original design using three wires.

If changing GDT, you could check Vge diodes while GDT is removed.  However, no way to know if zener diodes or TVS diodes are conducting.  Both are nominally 33V.

Another thought that I had was voltage sag caused by smallish bus capacitors (1000uF each) and a much higher MMC (.165uF vs .075uF) than designed?

You may get voltage sag.  I can't think of any resulting issue beyond reduced performance.

[FonziDaytona]

I hooked up my Bertan to the board and applied +500v (gradually) to each set of TVS diodes. Set one clamped at 435v and set two clamped at 425v. Guess they can be ruled out?

-Matt

[davekni]

I hooked up my Bertan to the board and applied +500v (gradually) to each set of TVS diodes. Set one clamped at 435v and set two clamped at 425v. Guess they can be ruled out?

Yes, presuming you biased the half-bridge supply.  If 0V across supply, then no way to determine which pair of TVS diodes is conducting, since they are bidirectional.

[FonziDaytona]

I put the hv power supply leads directly on each set of TVS diodes and then reversed the leads.

-Matt

[davekni]

I put the hv power supply leads directly on each set of TVS diodes and then reversed the leads.

If the TVS current at breakdown was sufficient to charge bulk capacitors, then this would be valid.

[FonziDaytona]

While I keep waiting for my replacement IGBTs, I had another question about the voltage drop and calculation for half bridges…Should bus voltage (400v approximately from voltage doubler) be half (200v)? So the calculation for energy in capacitor vs MMC energy would be (BUS/2)^2 instead of BUS^2?

-Matt

[Mads Barnkob]

While I keep waiting for my replacement IGBTs, I had another question about the voltage drop and calculation for half bridges…Should bus voltage (400v approximately from voltage doubler) be half (200v)? So the calculation for energy in capacitor vs MMC energy would be (BUS/2)^2 instead of BUS^2?

-Matt

Yes, you only have the energy available from one of the capacitors, at a time. Which you can also see is substantial as its the figure squared that you are halving.

[FonziDaytona]

Got my new IGBTs in (the C2D1) version which can handle another 100A.

Added a 7k resistor across the bridge output, and now I’m able to get things going with barely any voltage in from the variac. Here is a screen shot of current and voltage across the bridge. I am testing at very low BPS and on time with just the primary in place.

Edit: Excuse the scaling. I swapped my Pearson 110 CT for a triad CT (easier to fit in there) and forgot to change scaling around.

[davekni]

Added a 7k resistor across the bridge output, and now I’m able to get things going with barely any voltage in from the variac.

Glad resistor worked as expected.  Hope IGBTs survive now!  Scope captures look fine.

[thedoc298]

I see the on board MMC has bleeder resistors and my upgraded MMC doesn’t. Maybe startup IS the issue after all?

-Matt

This is what is tried to be fixed with a 10K resistor across the inverter output, to get rid of any residual charge that could confuse the logic at start of a pulse, not just start-up, but it would sound sputtering and unstable.

Did you take a look at all gate driver resistors, TVS and diodes from GDT to IGBTs? Many components can fail "half" at IGBT failures, like gate resistors going from 10R to 100R-10K and you get all these weird behaviours from differently damaged components in the drive chain.

Hello all, enjoying the chat, where (and I looked) would I find more info about the 10k resisitor and what problem it might fix. I know it's not necessary 10k. 

[FonziDaytona]

I see the on board MMC has bleeder resistors and my upgraded MMC doesn’t. Maybe startup IS the issue after all?

-Matt

This is what is tried to be fixed with a 10K resistor across the inverter output, to get rid of any residual charge that could confuse the logic at start of a pulse, not just start-up, but it would sound sputtering and unstable.

Did you take a look at all gate driver resistors, TVS and diodes from GDT to IGBTs? Many components can fail "half" at IGBT failures, like gate resistors going from 10R to 100R-10K and you get all these weird behaviours from differently damaged components in the drive chain.

Hello all, enjoying the chat, where (and I looked) would I find more info about the 10k resisitor and what problem it might fix. I know it's not necessary 10k.

It fixed issues where my coil wouldn’t oscillate at low power or act “weird” and then kill a IGBT. Lost a lot of them…

After adding a 7kohm resistor across the half bridge outputs, the coil starts up at very low input voltage to the bridge. No weirdness either.

Now back to tuning and seeing if I can push my IGBTs higher :-)

Here it is hitting around 21”

 

[davekni]

Hello all, enjoying the chat, where (and I looked) would I find more info about the 10k resisitor and what problem it might fix. I know it's not necessary 10k.

Here are several posts where I describe startup behavior and need for a resistor across bridge output:
    https://highvoltageforum.net/index.php?topic=1934.msg14466#msg14466
    https://highvoltageforum.net/index.php?topic=2330.msg17044#msg17044
    https://highvoltageforum.net/index.php?topic=2388.msg18296#msg18296
    https://highvoltageforum.net/index.php?topic=2624.msg19237#msg19237
    https://highvoltageforum.net/index.php?topic=1914.msg17230#msg17230
    https://highvoltageforum.net/index.php?topic=1350.msg9973#msg9973
If you still have questions, feel free to ask.

Now back to tuning and seeing if I can push my IGBTs higher :-)

Here it is hitting around 21”

Great to see it working!  Good luck with pushing just the right amount, not quite to IGBT failure.

[FonziDaytona]

Thanks again, David. That’s two of EVRs products that needed outside help and modifications to make them work.

-Matt

[FonziDaytona]

Managed to kill another IGBT. At least this time, I was pushing the coil near the max. No explosion, just sudden silence. I was near 400A, which seems low for these IGBTs. I’m thinking that this might be due to running 15v on the gates instead of something closer to 20v since the max pulsed rating is 300A.

Any reason not to modify the GDT to output 20v instead of 15v? I could wrap a 10 turn primary directly over the existing windings without much hassle…

-Matt

[thedoc298]

Hello all, enjoying the chat, where (and I looked) would I find more info about the 10k resisitor and what problem it might fix. I know it's not necessary 10k.

Here are several posts where I describe startup behavior and need for a resistor across bridge output:
    https://highvoltageforum.net/index.php?topic=1934.msg14466#msg14466
    https://highvoltageforum.net/index.php?topic=2330.msg17044#msg17044
    https://highvoltageforum.net/index.php?topic=2388.msg18296#msg18296
    https://highvoltageforum.net/index.php?topic=2624.msg19237#msg19237
    https://highvoltageforum.net/index.php?topic=1914.msg17230#msg17230
    https://highvoltageforum.net/index.php?topic=1350.msg9973#msg9973
If you still have questions, feel free to ask.

Now back to tuning and seeing if I can push my IGBTs higher :-)

Here it is hitting around 21”

Great to see it working!  Good luck with pushing just the right amount, not quite to IGBT failure.

Thanks a million, be hard going without you guys on this site.

[davekni]

Managed to kill another IGBT. At least this time, I was pushing the coil near the max. No explosion, just sudden silence. I was near 400A, which seems low for these IGBTs. I’m thinking that this might be due to running 15v on the gates instead of something closer to 20v since the max pulsed rating is 300A.

15Vge may contribute, along with hard off-to-on switching (no phase lead provided in this design).

Any reason not to modify the GDT to output 20v instead of 15v? I could wrap a 10 turn primary directly over the existing windings without much hassle…

Two down-sides to changing GDT winding.  One is that leakage inductance increases so Vge waveforms will have slower transitions and perhaps more overshoot/undershoot.  Another is that higher Vge can make hard off-to-on switching more stressful for IGBTs.  Increases switching current, forcing opposite diode reverse-recovery to occur faster.  When it does turn off, rapid end to higher current leads to higher Vce spikes.  Slower Vge edges can reduce this issue some, but also delay switching so normal primary current is a little higher by the time the off-to-on transition occurs.  Hard to say what the final trade-off will be.  The original design of EVR Microbrute appears to have 1:2:2 GDT winding, which was changed to 1:1:1 for later versions.
If you do change to a higher GDT ratio, I'd recommend dropping primary turns to 10 (per Microbrute instructions) to avoid adding even more leakage inductance.
(Higher GDT ratios also increase power dissipation in driver chips.  There may be margin to handle that.)
Another option is to trade out driver chips for versions capable of 20V and with direct FET outputs so GDT drive gets closer to full supply voltage (rather than Vbe less).  That would maintain fast Vge edges by keeping GDT 1:1.