MOSFETs often die when used within their rated specs

If you're only looking at datasheet specs like their rated voltage and current, then yes. However there is so much more information in the datasheet which is often overlooked. Perhaps most importantly:


The fact that a MOSFET can handle 500V at 100A, doesn't mean it will handle it however you like. There are limitations to how long and how often can a MOSFET take such abuse.

Thermal capabilities also cannot be overlooked! If you're using a large heatsink it doesn't necessarily mean that the chip temperature cannot reach over 150C.

In my experience; in hobbyist electronics MOSFETs die mostly due to people not paying sufficient attention to datasheet or wrong circuit design. Where as in consumer electronics, MOSFETs die most often due to poor thermal contact between the FET and heatsink.

Yes you can do it this way. Just make sure to properly rate the resistor for the voltage and power (I recommend highly overrate it). Because in case the resistor fails short circuit (which might be rare but it's possible), then the resistor will blow your bridge.

That's a very nice and clean build, good job!

I will try my best to answer your questions (more experienced coilers feel free to correct me):

Is my planned MMC suitable? Is it perhaps a bit oversized for this coil?
Is there a rule of thumb as to how much of a safety margin I should factor in compared to the DC ratings?

Both of these questions come down to how reliable you want your system to be.
If you want to show your coil to your friends once in a while, you can stick with DC rating. However if you want your coil to run reliably many times per month, I would at least go with the AC rating of the caps and perhaps even factor in some 30% safety margin.
The same goes for the Irms rating. I would say 146Arms is quite a lot for your coil, but hey, the more the better. Also with UD+ driver I assume you want to run longer on-times resulting in higher RMS current.

Do I connect my MMC as parallel series strings, or as a set of parallel caps in series that results in a grid-like structure?

Parallel series strings is the most often used configuration that is proven to work, I'd stick with that.

Do I need bleeder / balancing resistors for my MMC and if yes, what values / rating / type? Or is the 10k resistor across the bridge output enough?

It is known that UD1.3b driver was causing issues during the startup of each pulse due to residual charge on the MMC. This behaviour can be seen here:

I'm not sure if this problem persists with UD+ driver. However if you're going to use bleeder resistor, I would strongly advice to use resistors across MMC.
In case the resistors fails short circuit, it will destroy your bridge. If it's across MMC, it will only short out the MMC, and the bridge (nor anything else) should not be damaged.

Are the TVS useful, or do they increase risk for blowing my bridge in the case they overload and fail short? I saw conflicting information while looking around. Is it better to put two in series for 880V?

This question has been discussed a few times here already, I would say most people on this forum would agree that placing TVS diodes on the bridge only increases the risk of blowing up the bridge as it already happened to me.
The thing is that the bridge is very low impedance source and if TVS start to conduct, there will be high current flowing. Combined with the 880V of voltage drop, this is a huge power loss which will blow them up. And as semiconductors often fail short-circuit, they will short your bridge and destroy it. It's fine to place TVS on the Gates, as there certainly won't be potential kiloAmperes of current flowing and the voltage drop is only 30V. The bridge is a different story.

Also let's face it, if you're having 880V+ peaks on your bridge, chances are your coil is not well tuned and clamping those spikes is not the correct solution.

Is it an issue that one GDT ended up with one turn less than the other? I guess I will just have to scope the gate waveforms to check if there is a difference.

Most likely not, with 21V Gate drive, it should not matter that much. But scoping everything and testing for possible transistor heating is a good idea before you run it at full power.

I am thinking of cutting my heat sink into two parts since it seems a bit oversized, and I would end up with another heatsink for a future project. Do you think it is better to leave it as is? In that case I would put some stuff like the drivers etc on the free part.

Seems like you could be fine with half of your heatsink (if you're going to active cool it). However this depends on the power levels you want your coil to reach.

There is also the option of winding my secondary on a 160mm pipe instead of a 120mm one (same 0.3mm wire, same length). This would give me an aspect ratio of 1:4 and lower the frequency to ~80kHz, which would allow me to upgrade the IGBTs if I ever wanted (300A bricks). However looking at Mad's website, the resulting reactance of 65kOhm seems rather high compared to the recommended 50k-ish, and of course requires more space. Do you think it would be preferrable to go into that direction instead?

I think this 50k impedance doesn't matter really. I've seen coils running very well at much more off than 65k. Just make sure your coupling stays in the acceptable boundaries and that the distance between primary and secondary is enough to shield against primary strikes.
 

Yes, with one tiny change.  UD2.7 gate drivers output high, so low side NMOS FETs are on.  Otherwise exactly correct.

Thanks a lot for the confirmation

The DC blocking capacitor starts charged rather than at 0V.

Wouldn't a discharging resistor parallel to the cap solve this issue?

Just use the driver chip enable pins instead of AND gate on input.

I have thought of that, however IXDN6XX / IXDI6XX drivers do not use the enable pin. There is a IXDD6XX version of the non inverting driver that uses it, however there is no such option for the inverting chip. Also the IXDD6XX chip outputs High-Z instead of low when enable is low, that is not much helpful. I've picked these drivers a long time ago since they are in a TO220 package so they are easy to cool and they are not much costly, however now I see they are not the best in this situation.

Yes I've spent whole night on it, but I think I understand why the accident happens.

During the overtemperature event, the driver basically keeps the Gates floating, meaning one of the transistors will turn off slowly (through secondary GDT winding).
If the transistor don't switch off quick enough (before the primary current starts another cycle), the transistor will suffer heavy loss and will fail.

I've solved this problem by using another winding on the GDT, that shorts the transformer to GND, eliminating any residual charge on the transistors.



I'm not sure how this will work in practice and I still can't wrap my head around why an UD2.7C driver does not need such feature, to me the GDT drive circuit looks the same.

I'm not sure if the lack of response means that my question is a dumb one or if people generally don't have an opinion.

I think there are none dumb question and if there were, this would surely not be one of them. Sometimes you just need to give it a bit more time.

Is this a deal breaker

It is a huge deal breaker, those diodes carry the primary energy back to bus capacitors during ring-down. They basically act as a full bridge rectifier (assuming full bridge of transistors).

or can I just add external diodes across the collector/emitter junction

I've seen people doing this in SSTCs, however I see no reason why it should not work for DRSSTC as well. Just make sure to place the diodes as close as possible to the IGBT (ideally directly on the terminals of the IGBT)

I presume primary waveform is with scope ground (or - side of differential probe) to + line input rather than - (as used for waveforms in data sheet).

Yes the primary waveform was scoped relative to line +. I'm using a isolation transformer to power the circuit.

I'd guess core is saturating.

Well I have tried to change all the other components except the chip itself and transformer so it seems like it's either core saturation or faulty chip.

I thought these inductors with air gap are hard to saturate. I guess I'll have to try different transformer.


EDIT:
I've just tried replacing the transformer for a different one

Yet the results are completely the same, the chip starts cycling at the exact same load.
Thefore I think there is something bad with the chip, as I highly doubt 2 completely different transformers from 2 different circuit boards would have the exact same effect

I also wanted to know if I need to use TVS Diodes as I already have snubber capacitors.

In general I think TVS diodes are not a good idea. I've personally had an experience where a whole bridge blew because of them. The problem is that the the bridge itself is very low impedance, all the traces are thick and there is nothing that limits the current in case a TVS diode catches a voltage spike. Chances are the current is going to be very high, combine that with the huge forward voltage drop of the TVS diode, and you got yourself enormous power which the diode is not able to withstand not even for a brief duration of the spike. The diode will not survive and go short circuit, shorting the bridge and destroying everything.

I think the only part of a DRSSTC where TVS diodes can be used, is between the Gate and Collector of IGBTs. Fairly low currents of GDT drive combined with low voltage, makes it very unlikely that a short voltage spike destroys the diode.

I also think that if your bridge shows problematic voltage transients, a voltage clamping is not a solution. It's probably caused by some non ideal switching. Snubber capacitors or better design of the circuit is needed.

MMC - 2 strings in parallel of 2 in series of Cornell Dubilier 942C20P15K-F for 0.15uF at 2000VDC

Is that 2000VDC a typo? I found those caps to be individually rated for 2kV, giving 4kV for 2 in series. Anyway 4kVDC still seems quite low. I assume those bricks could be easily pushed to 500A-700A OCD, which would result to 4080V - 5700V AC on your capacitors at 130kHz. I would put at least 4 in series if your build needs to be reliable.

For low voltage testing, use one probe for triggering on some stable signal such as one driver output (one GDT input).  Move other probe from one signal to the next.

That's clever. Just recently I've programmed myself an application that takes an image from the scope as the input and converts it into text file. My scope can only export images...
I've used this app and some LaTeX to render this:


Only thing I didn't capture is the high-side Gate as for some reason at the time of measurement the duty cycle was a bit off when scoping the high side Gate, making the waveform unusable for this method.

It all seems to act pretty much the way I would imagine, except the primary and secondary current.

The secondary current was scoped on the feedback CT 1:20 with 1k burden resistor
The primary current was scoped on a CT 1:20 with 15ohm burden resistor

The low side and high side Drain-Source voltage has a different peak voltage, but that's just caused the way I measured it. (turned the variac off each time I put the probe elsewhere and then turned it back on, apparently I didn't manage to set it precisely where it was before).

Edit:
Here is the picture including the high side Gate-Source. It's only usable for the first cycle, then it's skewed too much.

It's interesting to see that the high side Gate voltage looks much "cleaner"