Hi Dave,

Thanks for sharing the stress test info on your C0G caps!  Fascinating that the ESR is going up without breakdown or loss of capacitance, makes me think some kind of electrode metal migration is happening, but this is certainly not the failure mode I would have guessed.

I've been using 0.1uF 500V C0G Kemet chip caps in assemblies of 48 sandwiched between copper plates, to make 4.8uF caps for induction heating use.  They seem to work great and handle 400A RMS at 110khz.  Much smaller than any film or other induction heating caps I've seen, though hard to say that its cost competitive with ebay deals or whatever.

I also used some 100pF 3kV C0G chips from TDK to make a MMC for my HFSSTC.  Seems to perform as well as high Q doorknob capacitors, though the little chips do run pretty hot.

Shorting MMC on a UD2 sounds potentially bad... pulse may be limited by GDT saturation.  However, i do think C0G MMCs are a good idea, especially if you have series connected units to avoid single event total failure.

Steve:  Interesting to know that you use external drain-source capacitance.  I'd wondered about that and ran many simulations both with and without external capacitance.  In simulation, I could get a little more power with the external capacitor.  It lowered primary resonant impedance, reducing how far out-of-tune it became as the secondary arc load increased.  However, I liked the simplicity and cleanliness of using internal FET Cds only.  FET current is close to sine wave.  No sudden changes in current at zero Vds.  Even though this FET has dual source leads, I liked the idea of no sudden current transfer, avoiding the resulting ring in the parasitic inductance loop from FET to external drain-source capacitor.

I had the same line of thought (about snubber current vs fet channel current), and for a period of time I suspected the added D-S capacitance was causing my circuit to self-oscillate in a worse way due to the interrupted current through the common source inductance (it blew up once and there was unusual ringing observed at the gate). However, in practice it seems to work fine despite a little extra ringing.  I try to keep the extra Cds about equal in charge to the fet's Qoss, effectively doubling the charge of the fet alone, all else being equal this should allow for twice the input current for the same peak voltage.  For the SiC JFET cascode that i use, the extra Cds should help reduce switch-off losses which are the dominant source of energy loss, i think, for most class E.  Other SiC Fet technology might not be as limited in switch-off speed, so perhaps you wouldn't see much gain from the extra Cds.  Also, i use the 4-lead package for lower source inductance.

Noticing a few things about your design - the secondary Fres is significantly higher than the operating frequency, and so I think this mode of operation, primary-resonant, is what provides graceful operation over wide power range with a fixed frequency.  Perhaps, like my HFSSTCs, once the arc capacitance is big enough to bring the secondary Fres to the actual operating frequency, the ZVS turn-on is lost.

Thanks for the waveforms, I also see similar higher frequency ringing in mine but not as pronounced on the drain voltage like that. I assumed in my case it might be ringing of the Cds internal vs snubber with any inductance between them, but you dont have that circuit... hmmm

Looking to offload some ferrites!  Perfect for that HV transformer, induction heater or power conversion project!  A few of the cores have cracked in my care (whoops!) and have been superglued back together, these will be discounted.  Many transformer designs will easily tolerate the extra air gap caused by the glue, but I digress...  The type of ferrite on these is generally not known, but for power ferrites, there's not a huge variation in types anyway.

Buyer pays shipping, USPS, UPS, FEDEX.  I accept paypal.  PM me if interested, willing to make a deal! 

Dimensions are per half-core section.

Large U: 93.5 wide x 76 tall x 30 deep(mm).  28mm x 30mm cross section with 36mm window between ferrite legs. 
I have 4 matched sets (plus 1 cracked set) and 1 of a different type of ferrite.  $15/set, $10/set for the cracked-leg one.


Large U/I: same large U above, I is 27mm wide, 30mm deep and 94mm long. Only have 1  $10/set


Wide U: 102 wide x 57 tall x 25 deep(mm).  25 x 25mm cross section with 51.5mm window between legs.  Only 1 set and one of the cores is broken/glued.  $10/set.


Long E: 70 wide x 54 tall x 32 deep(mm).  22x32mm cross section in the center and 11x32 cross section at the outside legs.  13mm window between legs. 5 sets available,  $10/set


Wide E: 80 wide, 39 tall, 20 deep (mm).  20x20 cross section with 20mm window between legs.  50ALL material (TSC ferrite),  4 sets available at $5/set.

Thanks for sharing this project!  Those long, ghostly, soft arcs are very nice for indoor demonstration!  Are you feeling the hair raise on your body, yet?  I put together a 5-stage (10X input multiplication) tower that could make 200kV and jump some 16" air gap, which agrees well with your 80cm for 400kV.  Even with 200kV, the ion streams could be intense, causing discharges several feet from the machine as "static" charge builds up on items nearby.

Is there any chance for recombining the parts from full-wave to half-wave to achieve higher voltages still? The high voltage DC corona plumes are fascinating, but very difficult to capture well on camera compared to the arcs, but at extremely high voltage it may be more interesting.

Thanks for all the interest, everyone.  I think the IGBTs have found new homes, my condolences to anyone who I could not supply with free IGBTs this time :-).

Hah! that website still exists!

I think Jimmy was the first to exploit double resonance h-bridge drive in a "transient" mode to make big sparks at high peak power, and thus the first recognizable DRSSTC.  That first machine ended up making up to 8 foot sparks with over 1kA from CM150s, really, quite groundbreaking for a first go. 

At around the time he was working on that first coil we were trading notes a lot.  That first DRSSTC used a microcontroller to generate gate pulses at a fixed frequency.  I was experimenting with "interrupted SSTCs" at that time where I began to stumble into near double resonance as my DC blocking capacitor started to resonate with some very low primary inductance, but it was Jimmy that helped me realize that and push it further.  I don't really think I invented anything in particular, but I did put out a number of decent designs that seemed to become standard. 

Like some sort of weird "high voltage forum fairytale", I actually work (remotely) with Jimmy, doing motor/inverter research and design. 

the largest drsstc
https://pbs.twimg.com/media/DrdddwEV4AARtqv.jpg

Not even DR, but yes, very big.  Lightning on Demand (Greg) can comment since he's the builder, but that machine has no resonant primary cap so its basically a huge SSTC.   

Nice switching there!

That looks like its ready to run with more space around it. 

Also, i noticed the mains line inductance was significant enough to require some decent bus , especially if your OCD just stops the oscillation at peak line current. 

Litz wire is often made from "solderable" magnet wire, which with a little abrasion while applying the iron will often succeed at wetting into the whole bundle of wires producing tap points which should only have a minor effect on the litz performance.

However, I'd agree with others that tuning the primary cap is a very practical way to avoid primary taps if you're careful about knowing your secondary parameters.  JavaTC is my go-to tool for designs. 

Ive built several QCWs with litz wire, and i will say it can make a significant difference in wire temperature (provided the litz is really spec'd for the high freq, your cooktop stuff might only be sized for 20-50khz, so its not as effective as the super fine stuff i used rated for 300-500khz), but generally the primary coil loss is relatively small compared to total energy, so you probably wont see any real boost in spark performance. 

I'm also building an IH (well, ive built a few now...) aimed at higher freq's in this range.  I'd be happy to share details, but its still in the works.  Also, as you already know, building these kinds of things are a steep learning curve with plenty of ways to go wrong. 

I'm using CREE SiC devices with a boot-strap type gate driver.  I have a post called "a three phase "tree phase" tesla coil" that has schematics attached for my half-bridge design, which im using the same design for the IH project.  I'm gonna try the C3M0065100K in full-bridge as im limited to about 12KW, which at 600VDC is only 20A or so (the switches might see 30A peak).   

The topology of the heater is a series resonant tank circuit fed by a step-down ferrite transformer with litz wire primary and water cooled secondary and a ratio of 35:1.  The bridge switching will happen strictly at or above the resonance frequency as this offers lower switching losses for the mosfets.  Power can also be controlled in this manner by operating off-resonance to control the boost that the resonance offers.

Alternatively, if the switching losses are too great, i will use a pulse-skipping method where the full-bridge can alternate between supply voltage at resonance, versus connecting both sides of the bridge to the same voltage, essentially putting zero voltage into the resonant circuit, letting the resonant current decay, before pumping it back up again.

I would say the SiC stuff is relatively easy to use if you are aware of a few things, like keeping gate drive parasitic inductances to an absolute minimum, and also watching the loop inductance of the half-bridge itself, because the higher switching rates available can more easily generate voltage transients.  The gate drive can oscillate and self-destruct if its sloppy (too much stray inductance).  I highly recommend the mosfets with a 4th "kelvin source" connection which eases the gate drive issues.

I've scored many CELEM capacitors on ebay recently, for pretty low prices.  Im using a pair of CSM150's of the .66uF type in parallel for this latest iteration. I do worry that some of them might be worn out, but the cost was small enough to take the risk.  The capacitance checks out OK on the 6 of them that i bought so far (i also got a pair of CP100, and a pair of CHF3 type 6). 

With all that being said, if you are in a hurry i would consider other options first as developing something like this is very challenging and likely to have a lot of hurdles.

I guess its worth saying why I want to build another IH, particularly with higher frequency capacity.  This heater is aimed at heating small copper pieces for purposes of selectively soldering them together as they are stacked with insulators and solder paste to form high-fill-factor motor windings.  The higher frequency will drive up the losses in my copper work pieces, getting it to soldering temperature even faster than my 210khz induction heater can now, which can take up to 60 seconds to achieve soldering temp, even with 500Apk circulating in my work coil.  Shorter heat time should limit the heat flow into neighboring coils and the stator core.