My QCW DRSSTC, small questions.

[Lucasww]

I've recently decided to build a QCWDRSSTC. I tested it for the first time today, and figured I should share it here with some things i've been wondering.

Specs:
UD2.9 Skip Pulse Driver
3.5 x 5 inch secondary, ~700T 34 AWG wire, 8x2 toroid, unloaded resonant freq around 300kHz
Upper pole operation around 370kHz
10nF 10kV MMC, 10 series 3 parallel B32642B0333J
Full bridge of FGH75T65SHD, Planned 150A maximum current
Async buck converter, with 2x FGH75T65SHD and 1x DSEI2X31-10B
100u buck inductor, 24u capacitance (doubles as decoupling on the bridge)

Pictures:

Inverter w/ FGH75T65SHD, capacitors and GDT


UD2.9 Skip Pulse driver (board by profdc9)


Ramp generator, using a modified version of Finn Hammer's design


Buck converter driver, using Loneoceans' design


Buck converter


And heres a ~2 foot spark, this was with basically no tuning, and a ~200v 7ms ramp


If I try to turn the voltage or pulsewidth up much further, I run out of energy in my bus caps. This makes sense, since I'm currently using 2 450v 3300u caps in series as a voltage doubler, as I live in 120v land
I have some smaller 200v 680u caps, and was wondering if theres anything wrong with using the smaller caps as a doubler, and putting the big caps in parallel on the output. Like this.


I know this will slow down the capacitor charging, but I don't plan on running more then 2 or 3 bps anyway for now, so if that's the only problem I'm okay with it.

If there as anything else I could do to improve it i'd like to hear, other than just tuning of course

[davekni]

I have some smaller 200v 680u caps, and was wondering if theres anything wrong with using the smaller caps as a doubler, and putting the big caps in parallel on the output. Like this.

That will work great.

[Lucasww]

Update, I tried it and it did work great! I now have it set up with 3x 3300u caps in parallel, and 2 680u caps for the doubler. I'm able to get close to 20ms 300v ramps before I get any voltage sagging, and make some very nice sparks.


Here's the bus cap setup on some aluminum bars.


Now, I'm gonna try and experiment with primary coils a bit. If i'm correct, Higher coupling should be better for QCW coils as long as theres no flashovers?

[All Science]

Great work on this coil! You've encouraged me to start working on my QCW again. Do you have anymore plans for this coil? Like putting it in an enclosure or squeezing a little more output from it?

[Lucasww]

Yes, I'm gonna try to do both. not sure how I'm going to do the enclosure yet, but I want it to look nice

[davekni]

If i'm correct, Higher coupling should be better for QCW coils as long as theres no flashovers?

Yes.  Higher coupling allows good power transfer over a wider frequency span, so performs better as arc capacitance reduces secondary frequency.

Higher coupling also separates upper and lower pole frequencies farther.  For a given uncoupled resonant frequency, higher coupling will increase upper pole frequency and reduce lower pole frequency.  Most QCW coils operate at upper pole frequency.  Increasing coupling without any other changes will increase frequency some.  Higher frequency is generally better for QCW performance, presuming your H-bridge can handle the increase.

[Lucasww]

Update, I did some very rough simulation in FEMM for some primary coils and came up with a design I liked, with a coupling around 0.44. I modeled some supports and had a friend 3D print them, and here's the result.





That's a 51" spark, 10x the secondary coil's length!
But, now I'm hitting OCD constantly at 150A. I'm not sure if my IGBTs can handle more, but they're rated for 150A continuous and are getting just warm enough for it to be noticeable after a minute at around 2BPS. Maybe they can take a bit more.
Alternatively, I might try lowering the current a bit. Maybe slightly lower coupling and going a turn lower on my primary? not sure.

[davekni]

Nice performance!

But, now I'm hitting OCD constantly at 150A. I'm not sure if my IGBTs can handle more, but they're rated for 150A continuous and are getting just warm enough for it to be noticeable after a minute at around 2BPS. Maybe they can take a bit more.

These IGBTs can likely handle more current, 200A or a bit more.  However, increasing current requires care:
First, phase lead and dead time need to be well adjusted.
Second, improved head dissipation during bursts.  Package immediately under die is likely getting much hotter during each burst.  To handle this, add a large copper heat spreader under each IGBT with good thermal compound between IGBT and spreader.  Apply mounting force over IGBT die rather than at screw holes.  Insulating pads between copper heat spreaders and heat sink.

[Lucasww]

First, phase lead and dead time need to be well adjusted.

I believe I have it pretty well adjusted already, but I've never used phase lead before so I'm not certain. I'll get some scope shots and send them here tomorrow if I have time, so you can take a look

Also, I tried going a tap lower on my primary, which seemed to ease up the current a bit while still running upper pole, but the frequency is now too low, sparks are shorter and more branched. I don't think theres a good way to do this without making a new secondary coil, but I think raising the resonant frequency of my secondary coil would help.

[Lucasww]

Okay, here's some scope shots of current and inverter output voltage. I had trouble avoiding noise on my scope but I think these look alright.



How's this look?

[davekni]

How's this look?

Looks great!  All you need is heat spreaders with mounting force over IGBT die and you'd be ready for higher current.

[tatemac]

Hi Lucas,

What's the temperature of your caps during or right after coil operation?

I'm designing a QCW too and am trying to figure out if my caps can handle the current. I looked up the data sheet and with 3P strings, that puts your MMC to be rated for about 9A RMS max at 100kHz. I was gonna use a string of 10 942C20P15K-F's which would be rated at 13A RMS, so I look forward to your answer. Thank you for any insight.

Best wishes,
Tate

[Lucasww]

Hey Tate, I have this graph from the datasheet of my caps, which states that the caps should be good for a bit over 5Arms at 300kHz.


Taken from here: https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/943/B32641B-B32642B.pdf

So 3p should be able to handle 15Arms. Either way, the caps have no noticeable heating during operation. I'm not running air over them or anything, although I do have them spaced out to allow for natural air circulation.

[Lucasww]

All you need is heat spreaders with mounting force over IGBT die and you'd be ready for higher current.

Dave, Is 1x1 inches of 1/8 copper plate enough for heat spreaders? if not, what size would you recommend? I have some general purpose thermal paste and 0.3mm silicone insulation pad I think should work fine.

[davekni]

Dave, Is 1x1 inches of 1/8 copper plate enough for heat spreaders?

That should be good.

I have some general purpose thermal paste

Any idea of the specifications for this paste?  The common old white paste was around 0.7 W/mK thermal conductivity if I recall correctly.  Good pastes today are 8 W/mK or higher.  Even 0.7 W/mK to a heat spreader is likely better than devices mounted directly to insulating pad without heat spreader.  However, I'd suggest finding some newer paste with good thermal conductivity.

and 0.3mm silicone insulation pad I think should work fine.

Probably good.  Any thermal conductivity specifications?  I presume this is a filled silicone intended for thermal conductivity, not a plain silicone rubber pad (which would be ineffective).

[Lucasww]

Any idea of the specifications for this paste?

My paste has a specified thermal conductivity of 1.6 W/mK, which I guess isn't amazing but not terrible.

Probably good.  Any thermal conductivity specifications?  I presume this is a filled silicone intended for thermal conductivity, not a plain silicone rubber pad (which would be ineffective).

It claims to be "fiberglass silicone thermal pad", with a thermal conductivity of 1 W/mK and breakdown voltage of 4.5kV. I've also heard that kapton tape can be used for this, presumably with thermal paste. Not sure if that works well though.

[davekni]

It claims to be "fiberglass silicone thermal pad", with a thermal conductivity of 1 W/mK and breakdown voltage of 4.5kV.

I'm guessing/hoping that this pad has some bit of silicone on either side of the fiberglass layer.  In other words, that the fiber glass reinforcement layer is thinner than the overall 0.3mm pad thickness.  Presuming so, that should provide compliance to fill in surface roughness of heat sink and copper spreader.  Otherwise thermal compound would be needed too.

I've also heard that kapton tape can be used for this, presumably with thermal paste. Not sure if that works well though.

Normal polyimide (Kapton) has relatively low thermal conductivity, around 0.12 W/mK.  Tape is usually thin compared to other thermal pads, which somewhat compensates for its low conductivity.  Would still be worse than your silicone pad at 0.3mm and 1W/mK.

Remember to apply sufficient clamping force over IGBT die.

BTW, there are much better thermal pads such as this for example:
    https://www.tglobaltechnology.com/product/tg-a9000-ultra-soft-thermal-pad/
This was just a random pick from a quick search.  There are many options.  I didn't look for any particular optimum selection.

[tatemac]

Either way, the caps have no noticeable heating during operation. I'm not running air over them or anything, although I do have them spaced out to allow for natural air circulation.

Thank you, Lucas!

[Lucasww]

Okay, I had to order some stuff from digikey anyway so I caved and ordered some better thermal pads and compound. Pads are 9 W/mK and paste is 8.7 W/mK. I'm also working on putting everything together into a nice enclosure. I'll send updates when that stuff gets here and i'm able to put it together.

[Lucasww]

Okay, Here's my new setup for the IGBTs. I have them on 1x1x1/8 copper blocks, held down with 3d-printed clamps, pushing down on the die. Is this good?

[davekni]

Okay, Here's my new setup for the IGBTs. I have them on 1x1x1/8 copper blocks, held down with 3d-printed clamps, pushing down on the die. Is this good?

Looks good.  My only concern would be that printed clamp plastic might soften at temperature or perhaps relax some even if not heated.  It is important to maintain force during operation.  Perhaps these clamps will do fine.  Or perhaps a metal spring could be added somewhere to provide compliance for any plastic creep.  Or change to a bent metal piece.

[Lucasww]

The heat is a fair concern, which I somehow didn't realize. Fortunately, I don't think it will be much of an issue. It's printed from PETG which shouldn't soften until above 80C. When running previously it was barely heating at all without any air circulation, I'm going to have a fan running over the heatsink now so I doubt it will even reach 40C. I'll keep my eye on it, though.

[Lucasww]

Okay, Heres an update.
I spent some time making a nice little enclosure for everything and put it together. Overall the results are good and the IGBTs don't get hot.




However, I'm having trouble getting arcs any longer than before. I decided to try making a new secondary with thicker wire, inspired by Loneoceans' QCW 1.5. the new coil with the same topload has an unloaded frequency around 370kHz with upper pole around 450kHz. As expected the arcs were a bit straighter, although unfortunately not as big. I messed around with coupling and toploads a bit and was able to get good results placing another topload on top of it, lowering the effect of spark loading.



Of course I want to use a single topload, and I don't really want to make a larger topload, Although I will if I need to. I've seen Steve Ward using a small MMC running through the secondary coil to add to the topload capacitance (here https://www.flickr.com/photos/kickermagnet/7383788136/)
I have a bunch of 82p 3kV C0G ceramic caps originally purchased for HFSSTC stuff and I'm wondering if it would be possible to run like 30 of them in series in parallel with the secondary coil. That should have a similar effect as adding the second topload did.

[davekni]

I have a bunch of 82p 3kV C0G ceramic caps originally purchased for HFSSTC stuff and I'm wondering if it would be possible to run like 30 of them in series in parallel with the secondary coil. That should have a similar effect as adding the second topload did.

Interesting idea.  Expect it will work.  My QCW top measurements indicate ~30kV RMS, so ~44kV peak.  30 x 3kV would be plenty safe for that.  Would be only 2.7pF total addition, however.  Perhaps two or three such strings in parallel would make a more useful difference.

[Lucasww]

Okay, I think i'm going to try that. Maybe 2x20 series would be better. Also, I've noticed the top of my secondary coil and underside of the topload get warm pretty quick. I'm guessing this is because of inductive heating? is this a problem?

[davekni]

I'm guessing this is because of inductive heating? is this a problem?

I'd also presume inductive heating.  If it gets hot enough to melt coil form or loosen winding or whatever, then it is obviously a problem.  Otherwise it is just a source of power loss that could otherwise go to increasing arc power.  Many TC designs include a gap between top end of secondary winding and top load to minimize eddy current losses (to provide a return path for secondary magnetic field).  My SSTC has a layer of plate ferrites between secondary top and top load.  That is not a normal solution.