Small UD3 Coil build log

[TMaxElectronics]

Hey everyone, I'm currently working on a small DRSSTC (or at least the box with the power electronics for that) and I thought I'd to a little summary of the progress here.

Why a small coil?
As you might have seen I currently only have a medium coil (2m) and no way of testing the software of my MidiBox (the next MidiStick version) without first finding some location, loading all the stuff into my car and driving out. All for maybe an hour or so of testing, which I then can't fully focus on anyway because there will always be somebody talking to me about the project
The solution is this project here. A small (-ish) coil, that I can easily run in my basement (and elsewhere).

Any requirements?
Most important thing is portability. I need to be able to easily carry the thing up the stairs and into my car should I want to make a presentation somewhere. And second it should be small enough to run in my basement without too many groundstrikes, while still giving a decent length arc.

In terms of power I just want to squeeze as much out of readily available parts as well as the space I have. I do however want to push ontimes as well as duty cycles quite high (600us, 20%) , so a little over designing is probably in order.

And then finally I want the coil to be very easy to assemble. My last coil has ended up being a little ratsnest of wires. I dred the day I kill an IGBT and have to take that thing apart again. So my goal is no wiring at all, except for the power input and the resonant circuit connections. Everything else should be on a PCB.

Current status
I have been noodling over some hardware layouts in CAD and probably settled on the one I will go with for the first prototype. I haven't been able to completely eliminate wiring (there is still an off-board 24V power supply), but I'm happy with how far I got it. Here are the specs for the power electronics box I settled on (haven't given much thought to the secondary yet):

Box Size: 300x300mm, 150mm tall
Driver: UD3 (I'll be making some upgrades to it in the process!)
Bridge: TO247 Full bridge, two in parallel for each position. I'll take whatever I can find (perhaps SiC in the future )
Bus: 4.000uF 600V (I have a bunch of 1.000uF 300V electrolytics around)
Supply: Single phase 16A, passive PFC (800uH inductor)
Resonant Cap: 100nF 6.3kV (dawncap, option for two in parallel)

PCBs are almost finished, they just need some verification before ordering. I've also started on some of those updates to the UD3 I was mentioning, I'll make a seperate post about that though.

Current design looks like this:


The bridge is hopefully designed for easy replacement of transistors when they blow up.

And the box will be made using my favourite build method: 20x20mm Aluminium profile extrusion and polycarbonate, should make the box nicely sturdy. I'll be releasing the CAD files for that once I'm done with it. In fact I want to make the entire coil design open source, so anybody could build one if they want to. I am kind of trying to use as few not off-the-shelf parts as I can for that too.

Let me know if you have any ideas for improvemets

[TMaxElectronics]

Hey finally some updates... yay

I finally finished the re-design of the UD3 (check out the UD3 thread if you are interested), allowing me to finalize the adapter PCBs too.

Overall cost is still quite high for the prototype, as I only ordered a small volume expecting to find some hardware bugs during the first few tests. In the end the entire coil will probably be around 500€ without any parts inventory on hand (so for me more like 200€ lol... hoarding ftw).

I think its time to start thinking about secondary size now
I'm aiming for a total height of something around 60-70cm. Not sure if I'll manage to hit that target though.

Here are some updated images:




If somebody is interested I might share the 3d cad drawings. If you have access to a cnc router or something similar it would be pretty easy to replicate the design.

[Mads Barnkob]

Wow! How did I miss this post?

Solid specifications and matching up fine for a cookie-cutter-coil, run of the mill of best practices!

The design is clean, renderings are really selling this coil If you think wires turned it into a non-serviceable rat nest, everything on a PCB is not necessarily easier to repair. But I get what you mean, all interconnections might as well be pin header connected.

The last rendering of the status LEDs, is that like a bar graph and with a full circle for highl-ightning the maximum reached?

[TMaxElectronics]

The last rendering of the status LEDs, is that like a bar graph and with a full circle for highl-ightning the maximum reached?

Uhhh very interesting idea. Right now they are just connected in parallel and hooked op to the UD3's LED outputs, but I will definetely keep that idea in mind

everything on a PCB is not necessarily easier to repair

True, but i was more aiming for a "I'm too lazy to run wires for a few days" kinda thing
Plus the PCBs are just plugged together with pinheaders as you said, so it can still be taken apart pretty easily.

[TMaxElectronics]

Oh hey, I just realised I totally forgot to update this log...

EDIT: I will be releasing the design files (pcbs, cad drawings etc.) to my github at some point in the future. I'll post about that here once I did. I also have some spare pcbs around, so if somebody wants to build one just send me a message)

Here's the current state of things:

I'm now on the second iteration of the pcbs in the coil, and they are working well now. The first had some stupid issues because I still keep forgetting to do a proper chekup on the schematic before starting board layout (so for example the precharge relay was connected to the precharge resistors with the nc contacts...).

I've also put in some work to fix the issues I had with the IGBTs blowing up as described in this thread here: https://highvoltageforum.net/index.php?topic=2502.0
I still need to search around for some bigger IGBTs as these already start desaturating at 150A, limiting maximum current to around 300A.
Although there is one good point to this: I have verified that the IGBTs are very easy to replace lol. Especially with the new revision of the boards, where you don't even need to unscrew the power board from the heatsink.

Another thing I have changed from the original design is the resonant capacitor. I initially planned to have a single 100nF dawncap/misc cheap aliexpress capacitor, but that turned out to not handle the current and also result in a resonant circuit with a very low impedance. At 350V VBus (rectified mains, with a bit of safety margin) the current would exceed the ocd threshold after just one and a half cycles. This resulted in peak currents far above the set pulse-skip threshold, further straining the IGBTs.
I now use two parallel 30nF caps, for a total 60nF resonant capacitance. Supposedly it can handle 120A now

Here are a couple of pictures of the assembled coil:

All in all I'm very happy with the result, although the polycarbonate I'm using for the case is just so incredibly easy to scratch

The secondary resonator is now 90mm pipe with 0.2mm wire and a 100mmx450mm toroid. This is the part I am least happy with though... Fres is around 150kHz, which is much lower than anticipated. I had to connect the primary at the absolute outermost point and  am still tuned slightly too high. No idea why javatc was off by such a large amount. That is probably also the reason fo the slightly disappointing performance. I'm also really starting to dislike the aluminium air-duct toploads... they are just sooo fragile and get dented pretty much from just looking at them the wrong way. I'll definitely be changing this over to one made from bent aluminium tubing in the future. I'll also hopefully make it disassemblable for easier transporting.
There is also one other problem with the secondary, but I'll get to that in the performance/thermal section of this post

Here are the parameters I settled on for now:
Ontime: 1ms (might go higher, it sounds just so awesome )
Primary current: ocd set at 165A, current peaks around 270-290A at 350V
Max dutycycle: 15% long term, midistick compressor set to limit to 25% short term (~500ms)

The coil makes around 0.8m long streamers, from a 0.3m long secondary. Not too bad but could certainly be much better with improved tuning once I fixed the secondary.

Thermals:
This is the point that I'm the happiest about. The entire power electronic section is still nowhere near its performance limit. For verification I glued one of the ntc's directly to the IGBT case, to at least give me somewhat of a better insight on the temperature of the transistors.

You can also see the aluminium plate heat-spreaders that I added for each pair of paralleled IGBTs to increase the surface area contact with the insulator sheet.
Here are the temps that the different parts of the coil got to after ~15min of playing music. Input Current was peaking at 20A from 230V with the average probably around 10-15A (so perhaps 2KW avg and 4KW peak, didn't measure the power-factor). Ambient temp was probably around 15-20°C

Heatsink: 30°C
IGBT tCase: 45°C (highest peak I saw) ~40°C avg
the following were measured with my thermal camera, so not that accurate:
Bus caps: ~ambient temp
Resonant caps: 30-40°C
Wires: 50°C
Primary: no idea, but not too hot to touch, so <60°
Secondary: 80°C+, waaay too hot to touch

So almost all good, except for the secondary... that got so hot that the wire separated from the pipe just from thermal expansion. Especially the the ends of the wire that went into the pipe were bad, as they even started slightly charring the pvc.
I think I'll definitely want to go to 0.25mm or even 0.3mm wire. That would also fix the issues I was having with the resonant frequency being too low.

And finally a bit of footage from the coil, including music after the static tones at the beginning (still uploading at the time of me writing this):

[flyingperson23]

That sounds really good, I might have to blow up my coil trying those on times. Is this using the standard midi stick, or the super secret updated version?

I'm not sure if this has been suggested or if it's a bad idea, but would it be possible to attach the secondary in a way that leaves the top and bottom of the pipe open, then blow air through it to provide active secondary cooling?

[davekni]

That sounds really good

Agree.  Sounds great!

Primary current: ocd set at 165A, current peaks around 270-290A at 350V

How are you getting 270-290A with OCD set to 165A?  Does current rise that much in one cycle?

The secondary resonator is now 90mm pipe with 0.2mm wire and a 100mmx450mm toroid. This is the part I am least happy with though... Fres is around 150kHz, which is much lower than anticipated. I had to connect the primary at the absolute outermost point and  am still tuned slightly too high. No idea why javatc was off by such a large amount.

What did JavaTC predict?  I haven't seen JavaTC be far off myself.  Is 150kHz secondary resonant frequency by itself or lower pole of the primary/secondary system?  Did you measure secondary inductance to compare with JavaTC?  Even if you don't have an LRC meter, connect a known-value capacitor from secondary top to ground (w/o top load in place), measure resonant frequency (signal source to secondary bottom), and calculate inductance.  Use a capacitor well larger than winding capacitance, 1-10nF is good.  If low frequency inductance matches JavaTC, then issue is with capacitance.

Secondary: 80°C+, waaay too hot to touch

Do you know with reasonable certainty the secondary coil form material?  Some plastics, especially filled ones, can have high loss and high dielectric constant.  Could possibly explain excess heat and low secondary resonant frequency.  (I learned the hard way on my low-frequency QCW coil that epoxy resin is very lossy and fairly high dielectric constant.  Normal use for coating is thin enough to cause little degradation.  Much thicker inside layer to insulate internal primary was problematic for me.)

I'm also really starting to dislike the aluminium air-duct toploads... they are just sooo fragile and get dented pretty much from just looking at them the wrong way.

A simple way to make aluminum air duct somewhat more robust is to keep the inner diameter completely compressed.  Expand only the outer diameter enough to accommodate curvature.  Requires more duct material, but is quite a bit more robust than when stretched out closer to maximum length.

IGBT tCase: 45°C (highest peak I saw) ~40°C avg

Thermal camera would likely be more accurate for IGBT case temperature than lug-mounted thermistor.  We've used such lug thermistors at work.  Thermal conductivity from thermistor to wire leads is similar to conductivity to lug.  Measurement is more like an average between wire temperature and device temperature.  Also responds much more slowly than a thermal camera.

I'm not sure if this has been suggested or if it's a bad idea, but would it be possible to attach the secondary in a way that leaves the top and bottom of the pipe open, then blow air through it to provide active secondary cooling?

Seems quite possible.  I'm considering this for my CW SSTC.  Temperature rise at its existing 3kW is OK.  Planning to make new drive electronics and primary coil for running it at 8-10kW, which may cause problematic temperature rise.

Impressive mechanical construction!  Much better than anything I'm capable of

[TMaxElectronics]

Do you know with reasonable certainty the secondary coil form material?

I doubt that is the issue, the pipe is polypropylene
Outside is epoxy coated, but only very thinly, so not sure if that could be the cause. I measured the thermal impedance to ambient air at around 0.9°C/W, which means a total power output of over 70W to reach the 90°+ that it did during the test run.

Air cooling is an interesting idea, although I'd be concerned with potential contamination of the inside of the pipe. I've had internal arcing issues with my coils in the past from simply not wiping the inside of the pipe down before winding the coil. A fan blowing (dusty) air through it might deposit enough to cause problems again. Perhaps its time to finally build a freon-submersion cool secondary coil

I think I'll want to reduce the power dissipation by at least half, which would require 0.3mm wire if 100% of the losses were resistive. The entire wire is also still within the skin depth, so the size increase should make a significant difference in the power dissipated.

Thermal camera would likely be more accurate for IGBT case temperature than lug-mounted thermistor.

True, but since I can't run the coil in its disassembled state easily this should be enough for me

How are you getting 270-290A with OCD set to 165A?  Does current rise that much in one cycle?

yes, and that is despite the reduced resonant cap size. Matches up with the calculated amount though

Impressive mechanical construction!  Much better than anything I'm capable of

Thank you

[davekni]

I doubt that is the issue, the pipe is polypropylene

Yes, polypropylene should be very low dielectric loss.  I haven't heard of PP being blended with fillers, so that seems unlikely.  The only down-side to PP is low surface energy.  Epoxy or other coatings typically don't adhere well.  Plasma cleaning is often used to activate the surface (break surface bonds) to make adhesives work with PP.  That could be contributing to delamination when it gets hot.  Of course, reducing temperature is good for other reasons too.

I measured the thermal impedance to ambient air at around 0.9°C/W, which means a total power output of over 70W to reach the 90°+ that it did during the test run.

70W sounds reasonable for your high duty cycle.

Air cooling is an interesting idea, although I'd be concerned with potential contamination of the inside of the pipe. I've had internal arcing issues with my coils in the past from simply not wiping the inside of the pipe down before winding the coil.

I wonder what coating or contamination was on the pipe you needed to wipe down?  I wouldn't have thought typical environmental dust to be that much of a problem.  Perhaps it really is.  I have little experience with such.

yes, and that is despite the reduced resonant cap size. Matches up with the calculated amount though

Yes, makes sense after looking back at your 150kHz frequency and 60nF MMC, 17.7 ohms Xc.  350V * 4 / 17.7 = 79 amps per cycle peak primary current growth rate w/o considering secondary.  Energy cycling back from secondary must be providing the occasional additional current to get from 165A to 270-290A in one cycle.

I'm guessing performance would be better with a higher primary impedance, perhaps 2x or at least 1.5x (30-40nF and corresponding higher primary inductance).  With less current change per cycle, average within-pulse current could be higher for a given max current.  Either that or higher-current IGBTs to match your low impedance primary.

[TMaxElectronics]

That could be contributing to delamination when it gets hot

Absolutely, although another big issue is the epoxy not even getting to the pipe's surface... with the thin wires it just doesn't seem to pull itself inbetween the windings as it did with my large coil.

I'm guessing performance would be better with a higher primary impedance, perhaps 2x or at least 1.5x (30-40nF and corresponding higher primary inductance). 

I do use two parallel 30nF resonant caps, so for experimentation I could just remove one of those
The current primary is as big as I can get it in the footprint that I am targeting. I might have to go to a double helix primary in the future, which would certainly be interesting to try too.
But for now I'll stick with the primary I have and work on a new secondary and topload first

[TMaxElectronics]

Ok so I just tested the new secondary. Wire diameter is now 0.3mm, which pushes fRes up to 215kHz which is a little far but I'm probably going to make a slightly fatter topload. Currently I'm just using one I made ages ago for what was to be a "crazy powerful coil". I guess that statement kinda evolved a bit
The secondary still gets very hot however... I'll get to that later

I've also upgraded the IGBTs to the highest current ones I found for a price I was willing to pay: FGH75T65SQD
The datasheet output characteristic plot only goes to 300A, but especially at 19V on the gate it seems like they can handle MUCH more. Even at 175° and 300A per IGBT I'd say desaturation isn't a problem.

When re-tuning the coil I went for absolute lowest possible fres that would still give me a streamer, and coupling is at around 0.12 now (according to javatc).

There is now another problem though... the power source
I haven't gone past 400A ocd yet (~500A peak) still at 1ms ontime and 15% maximum dutycycle (short term max is 23%) and the coil already draws 15-20kw. That is from a single phase 230V supply that is rated at only 16A. Peak current my clampmeter showed was 117,6A. I honestly think the coil would do 2m streamers if I had a ceiling high enough and more power. Thats also why the secondary gets hot again 

This vid has the catcher rod 1.5m away from the breakout:

Not a valid youtube URL Hmm embed doesn't seem to work... here's the link: https://youtube.com/shorts/LmHMwYoh9tM

The sudden end is from the breaker tripping (and no, not because I killed any IGBTs this time )

Guess I'll have to upgrade to three phases next. Although that is going to be pretty annoying with only 650V IGBTs (rectified mains is up to 600V here).

[davekni]

Impressive performance for this small coil!

I've also upgraded the IGBTs to the highest current ones I found for a price I was willing to pay: FGH75T65SQD
The datasheet output characteristic plot only goes to 300A, but especially at 19V on the gate it seems like they can handle MUCH more. Even at 175° and 300A per IGBT I'd say desaturation isn't a problem.

I'd agree that desaturation is unlikely to be an issue.  Transient thermal impedance is more likely limiting factor, especially with 1ms on times.  These are nice fast IGBTs.  As normal, fast comes with the trade-off of increased Vce on-state voltage drop.

[TMaxElectronics]

Impressive performance for this small coil!

Thanks

Transient thermal impedance is more likely limiting factor, especially with 1ms on times.

I agree, I'd need to do the maths again but I did test it at a very low repition rate (5Hz or so) at up to 2ms with no problems. Apart from the power consumption again lol, the 2ms test at 400A ocd made the analog gauge of my variac jump to 10A every pulse despite them being so far apart.

I'm now kind of thinking about going to yet another resonantor design. One that has a much higher fRes and then going qcw. I think the bridge should be able to handle it jsut fine

[TMaxElectronics]

Oh hey what a surprise, the project isn't actually dead or finished

I've been thinking about a collapsable topload design for a while now but not come up with anything great yet. Has anybody perhaps already built something like that before? I'm just constantly annoyed when having to carry around that tube-toroid that gets dented when you just look at it to strongly. A full sceletal one would of course be a nice alternative but transporting that is really almost as inconvenient as the current one

And probably the bigger news: I've started the process of documenting and open sourcing this entire project, so others can copy it if they want to. So far I've only gotten the three main PCBs (plus of course the UD3) ready for publishing but the rest is still to come
I'll also be including some short build instructions along with it. Perhaps not enough for someone to build this as their first coil but certainly enough for some experienced person to do so.
Here's the repo: https://github.com/TMaxElectronics/Mini-Coil/