I planned on making a thread about my DRSSTC for a long time but I always postponed it because I still had to fix this and change that. Now, after it exploded on me for the third time I finally write this post. While it's still kind of a presentation it's mainly because I have not the slightest idea of what's going on and hope you guys can help me. But be warned, to give you a complete picture, I got to go back in time a bit because I've had issues with this setup since first light - which was 4 years ago.
To keep the length of this post somewhat reasonable I only link to the pictures and videos instead of including them within the post. While I never talked about the build process here, I documented everything in the German equivalent of this forum (mosfetkiller forum).

Just to give you an idea even if you're only skim reading, here's one pic/video (the one of the recent test and failure):

Current specs

• When designing the coil I aimed for 32A 400V 3 phase operation.
• Filtering: 1,5mH frequency drive sine filter on each phase. Boosts the idle voltage from 565V to ~650V.
• Rectifier: 3x 35A 1kVdc fullwave modules, only using "half" of each one.
• Full bridge made of 4x SKM200GB128 300A 1200V half bridge modules paralleled in pairs
• Laminated bus bar design. Multiple layers of 2mm aluminium and silicone sheets. held together by Nylon M6 screws.
• 10x 3300uF 450V Epcos bus capacitors (measuring 3000uF and 20mOhm ESR). 5 parallel "strings" of 2 in series for ~8000uF @900V and 250Arms
• 4x 680nF snubber capacitors distributed across the bridge
• MMC: 14x 5uF 1200V Arcotronics GTO snubber for 360nF @12kV, 1.6kApk, ~160Arms
• Primary: ~5 turns, made from 10mm copper pipe
• 10k resistor across primary tank
• Secondary: 20cm x 105cm, ~1700 turns, 0,6mm
• Topload: 9x 25cm balls arranged as ~1m "toroid"
• GDT: ~60x40x20mm (outer dia. x inner dia. x height) N30. 11:11:11:11:11 winding/ratio. 1 GDT per pair of halfbridge.
• Gates: About 13nF gate capacitance, 5R1+1N5819 antiparallel diode, 24V bidirectional TVS diode. Gate signals show ~600ns/400ns rise/fall time. Only one overshoot clipped to 24V, stable plateau at 22V
• Driver: UD+, set for 900A pulse skipping.

The issue
After the most recent events it seems to me like all the issues I've had over the years could be explained by one root cause: serious overvoltage on the bus. And with serious I mean "visible 5mm blue arcing" overvoltage. However I don't have the slightest idea where that would come from.

History of events
When I started with this project 5 years ago my only power electronics experience was an SSTC which I never really finished and never worked realiably. I had read a lot before starting but tbh I didn't know half as much as I do today.

About two years ago I made a partial redesign of the setup to fix a couple issues I had encountered til then. Here are the important differences compared to the current setup:

• GDTs made from CAT5 patch cable as I had seen this being recommended at various places. Of each twisted wire pair I took the white cable and connected all of them together to form the primary. Every colored cable was connected to one gate.
• No input filter or choke
• Different bus bars with laser cut polycarbonate isolation. Creepage distance of 2mm. An early CAD model can be seen here.
• The bus capacitors were all paralleled, meaning not 5 strings of 2 but 2 groups of 5 parallel.
  
• 2x 440V TVS diodes in series across each IGBT pair (mounted inbetween the modules, visible in the pic linked at point 2 below)
• MMC made of 15 (instead of 14) GTO snubbers for only 330nF instead of 360nF
• Thinner wiring in the primary circuit

Pictures of the old setup

And finally, the list of important events/failures over the years:

• Popping and crackling noise from the bridge. Solved by cleaning off the burns from the laser cut isolation between the bus bars that likely caused small arc overs.
• Explosion of an IGBT module when first trying 3 phase (400V) operation (static load, very low power). Exact cause undetermined. I found signs of arc over between the power rails.
• Redesign. I spent a lot of time designing this in CAD and thinking about how it would be assembled. Turned out I was wrong and the screws cannot be reached. Don't ask me how I assembled it (and disassembled when it went boom). Let's just say it took hours.
• Burnt GDT. When taking out the GDT I found that it had burned through the isolation where it touched the grounded heatsink. Dismantling it I found another spot where the isolation of one primary-secondary twisted wire pair had molten away - only <1mm of air was left as isolation. No clue why this happened. Rewound it with individual wires and potted afterwards.
• Power run on 3 phases with the new setup and a static load, measuring temperatures and other things. Went fine for about 15 minutes. Then, when refiring it after an hour of cooling down, an arc over between positive and negative rails occured. Damage: One highside and one lowside IGBT in one halfbridge pair failed, one of the modules had a crack in the case. Both modules got replaced. Steve Ward thought it was likely that one IGBT failed first and the resulting flash would have shortened the rails. He also was the first one to bring up the insufficient camping force of the Nylon screws. For some reason neither me nor anyone else previously thought about that. I somewhat doubted that the IGBT failed first. Why would it? It didn't fail during the high power run. I saw once again spotweld marks on the rails and came up with the suspicion that the Nylon screws - especially after having heated up and cooled down - were insufficient for a good contact between bus bars and IGBT terminals, causing a spark which in turns caused the arc over between the power rails. As a temporary solution I added another piece of isolation such that no spark from one terminal could reach another one.
• Another run of the coil (with secondary) on 400V went fine. Gave my best to tune it but the output was not great. This is another issue; since the redesign I haven't had the same output as before. Previously I got bright >2m arcs with 230V 16A single phase input. Now I needed 400V three phase for a similar output (though 3 phase is notably louder than single phase - even with less output).
• First test run (with secondary) with my Syntherrupter. All events cited above had been done with its precursor (no polyphony, no special effects, so not much to go wrong). The test run went rather well until the coil once again went boom. This time - to my surprise - the arc over occured between one of the rectifier terminals and its case. I'm pretty sure that it's not the rectifier itself that failed first because the potting shows absolutely no signs of cracking as you'd expect it if >1kJ would go though it. The terminals and the case however show clear burn marks. Nonetheless, the rectifier was dead. I couldn't explain myself this event. The previous theory can't explain it and I couldn't see how an arc over at this place could occur without any IGBT breaking down.
• I rewired the rectifiers (see quick specs, I have enough redundancy for such a failure) and slowly restarted testing (on 230V single phase though). The output was not consistent; I can only describe it as "stuttering" (I can upload a video of it if someone's interested). Increasing the power did not really help. I searched for other issues but couldn't spot anything. To get an idea I hooked up my scope to the bridge outputs (1 chn per output). Without the ability to scope the interrupter signal or primary current (didn't have the necessary parts on hand) The scope shot was pretty useless as I had nothing to trigger on. However, the scope did make another big difference: the issue was now gone. It even seemed to me as if I'd be getting more output now than before the bang (more similar to what it used to be before the redesign. I have no explanation for this but after a bit more testing I ended it there.
• Few days later I was supposed to do a demo of the coil. However, because of bad weather and other reasons it should be a static load only (using it as improvised induction stove; we needed one). Again I heard that same stuttering, but this time - without the loud arcs - I could hear that it actually sounded like arcing around the bus bars. And after looking around for a while I found that indeed I could see it arcing between positive and negative bus bars. Small blue arcs that matched the fizzling/stuttering sound. Since I reacted immediately when I saw it I failed to take a picture or video of it and can't tell you whether it was inside or outside or through the silicone isolation. I was absolutely speachless. Why the heck are there even such voltages?? How comes that it doesn't cause a short? How do the IGBTs survive that? With easily 5mm long arcs all the previous event's aren't surprising anymore (though I never noted such fizzling/stuttering except for the very first point of this list).

And here we are. Sure enough the Nylon screws are a known issue and tbh I was expecting another failure related to them but didn't care for the time being. I do have a new design ready to manufacture (with steel screws and easy assembly) but til I'd find the time to actually build it I didn't mind risking another crash (maybe because I have >20 replacement IGBTs...).
Anyways, I cannot see how high contact resistance would cause such overvoltages. Neither can I think of any other reason for that. Why does a scope with 10MOhm probes affect any of all this? Also, why did it seemingly increase the output? These recent experiences make me feel like I'm missing some important knowledge when it comes to DRSSTCs... So I hope you guys got some ideas.


Thanks for reading!
Max

Hi everyone,


As I've written in other topics, I'm using a software called "SynthFont" for playing MIDI files with tesla coils. Usually, no matter what software you use, you need some kind of converter*. Either loopMIDI and hairless MIDI<->Serial bridge, or a USB to MIDI cable which includes a USB MIDI to serial converter.
However, after a lot of testing and refining, the developper of SynthFont integrated support for direct serial out. So you can directly from his software send any MIDI channels to any serial COM port.

SynthFont is entirely free, supports sound fonts (important for good audio quality), and allows you to view and edit the MIDI data. It looks a little old but so far I haven't found another free software which is as easy to use. There is also SynthFont 2, which has a few additional features (like VST plugin support) but its not free. For playing MIDIs with tesla coils the free version is absolutely sufficient.
SynthFont Download: http://www.synthfont.com/
Best sound font I know: https://sites.google.com/site/soundfonts4u/#h.p_biJ8J359lC5W

What setups do you use?

*If you happen to have a "MIDIStick" from TMaxElectronics you do not need to convert to serial because it natively supports the USB MIDI standard. However that's the only interrupter with this feature AFAIK.


Kind regards,
Max

Hello everyone!


I'm writing a rather big FAQ on DRSSTCs (but also covering other TC-related topics) in another Forum, the mosfetkiller forum. It actually is the "german 4hv forum"; simply the biggest community of electronic, hv, etc. interested german speaking people. As I'm way better in german than in english, I'm mostly writing over there, which also is the reason, why my FAQ is in german.

The reason I post it here, is mainly because questions I'm trying to answer with that FAQ also appear over here, and I think it might contain some valuable informations (although it's still a huge construction site). A short check prooved, that google translator doesn't do a bad job, so I hope it's useful here, too.

In no way do I want to "compete" against Mads and his fantastic DRSSTC design guide. I have a different objective in mind. While his work contains a lot of practical tips, how to design it, which means, that you already have to knoew what exactly you're designing, I'm trying to focus on the understanding behind it; how to get there. People should not only know how to build a DRSSTC, but also how it works, and why they should do things the way they do. This includes for example, why a DRSSTC can't be driven by an oscillator or by secondary feedback.
Therefore I'm starting with very basic things, so that really everyone has a good (well, hopefully good) entry-point to this stuff. Building up step by step on what's already been explained, there shall eventually be no "black box" left.

I know that there's still more work to be done then work that has already been done. Also, there are some small errors around. Therefore any kind of feedback, tips, ideas would be highly appreciated. Once it gets to a more advanced stage, I'll definetly consider an english translation - if there's any interest
For now the roadmap is to finish the (technical) DRSSTC part (40% done?), then I want to write another chapter about safety (again (starting with) explanining rather basic things like why you can't connect your scope immediately to your bridge's output) and finally there might also come a chapter about testing, building up on the safety and the theory chapter. But as this gets pretty practical (and therefore potentially covered by other great guides in the internet), I'm not yet sure about that.

If you want, I can keep you updated about future changes.


Kind regards,
Max