First DRSSTC, Full Bridge PCB & IGBT Selection question.

[Beggernator.]

 Hi.
First time post from a long time lurker here.

As title suggests first DRSSTC build. Plan is to get it running as a SSTC then add in a MMC and tweak the control board to get it operating as a DRSSTC.

In the past I have read a lot of Steve wards older content that is available on his site. Plus all of the available Kaiser power electronics blogs and guides.

More recently however I discovered the great detail on loneoceans site. And from that gained the inspiration to do a build that targets a 2kw coil running on a 3 inch secondary.

From most of the searching I have done nobody seems to be using cutting edge silicon in there coils, IGBT selection is an artform in itself without considering the use of newer Silicon Carbide 1200v MOSFETS.

I decided to go for a halfway house. some modern new to market IGBTs driven with tried and tested GDT setup.
The ones I settled on are IKZA50N65EH7 from Infineon datasheet : https://www.mouser.co.uk/datasheet/2/196/Infineon_IKZA50N65EH7_DataSheet_v01_10_EN-3421412.pdf

They have comparable switching times to the frequently used IRGP50B60PD1 Warp 2 IGBTs, while having a CC rating of 80A and a datasheet repetitive pulse rating of 200A.
In addition to this they have separate emitter connection for the power and the gate so drive noise should theoretically be lower.

So I suppose its a very wordy way of asking do people think the IKZA50N65EH7 is a good fit for coil use. There are higher current versions of the same family but switching times get significantly longer. possibly a good upgrade option for a larger, lower frequency coil?

And my second question was, Could I get some feedback on the PCB layout I have done for the H Bridge. Its taken heavy inspiration from loneoceans 80mm fulbridge. I have made the board slightly bigger. increased capacitance to be inline with the expected higher current, and lower frequency design target.

Board will be made with 2oz copper, ideally I would like more but didn't fancy tripling the production costs.

Lastly, many thanks to all those who have posted content that enables people like me to learn about this fascinating branch of physics.

[Mads Barnkob]

Welcome to HVF

TO-247 packages only have so much thermal mass, that in my experience (older tech like 60N60) can not be pushed beyond 250-300A peak. Either you calculate your heat dissipation losses and make up if you want long life time or long sparks. Maybe just double up on them, 2 IGBTs in parallel per leg. Like the PCBs you can find in the UD2.9 drivers pack from Prof9DC in the sticky thread.

Feedback on PCB: INCREASE all creepage distances. There is no reason for tight tolerances. With high frequency high voltage present, things just behave differently. Might as well make it harder for that flash-over to happen.

SiC and GaN are far far superior, as also shown in the threads like Anders Mikkelsen. But they are more complex to drive. You can not use GDT and that makes 90% of people go for IGBTs.

[Anders Mikkelsen]

H7 IGBTs are nice, these are pretty new technology. From the datasheet, it looks like switching losses are comparable to IRGP50B60PD1 for the part you mentioned, while conduction losses are less than half, a significant improvement hopefully giving better overcurrent handling and margins. The kelvin source is also a nice upgrade, at least for hard-switched applications like PSFB QCW.

Layout looks good and low inductance. I'd remove the copper under the bleeder resistor, as the electrical isolation of resistor packages is often not great. Precharge is also recommended, though this can be added externally with a resistor in series with the supply and a relay to shunt it. I would use thinner traces for the gate drivers (a few tens of milliohms of trace resistance doesn't matter here) and route them closer together to reduce the inductance. I'd also consider removing the switching node plane under the GDT primaries to avoid coupling transients into your gate driver outputs during switching, this should still leave enough copper to handle both the peak and RMS currents in question. I recommend going with 2 oz copper (70 µm) if it's not too expensive, but it's probably not strictly required here.

SiC transistors perform well, but the advantages for short-pulse DRSSTCs are not great, since heating mainly goes with I^2 compared to directly proportional with I for IGBTs. For higher frequency and more CW-like coils, they can be of great benefit. Gate drive can be pretty simple if using cascode parts from USCi for example, as these will accept standard symmetrical gate drive voltages. For pure SiC MOSFETs, the gate drive requirements are asymmetrical, but there are plenty of chips that can drive them directly in this case. So the comparison is between two gate drive chips and one or two GDTs, vs. a gate drive chip and a few isolated power supplies. Neither is much more complicated than the other, and the GDIC solution provides the bonus of robust UVLO per gate channel and dead-time insertion, so I personally would use these even if making a coil with IGBTs these days. But the power supplies do add some cost unless bought in bulk, and there are limited documented projects using these, which is what I have been trying to address with my latest builds. Many people have had bad experiences with older generation junction-isolation halfbridge drivers, and these are very finicky to use, but the newer fully isolated parts solve all these issues.

[Beggernator.]

Many thanks for the reply's gents.

Good to finally be chatting about ideas instead of just reading others.

Thanks Mads for the spot on the clearances. I based them on 1.5mm for 400vDC but forgot the primary output might ring up much higher then that. will increase all of the tolerances as is still plenty of copper.

300A peak sounds good for a first coil. can point a 50cfm server fan at the heatsink and see what they will do first. I did see the concept of parallel bridge setup on some of Gao's posts.
As Anders points out with nearly half the conduction losses, and decently quick switching speeds compared to other market options I have a morbid fascination to see just how far a single one could be pushed.

And on the note of Anders it would appear I have some after dinner reading to do. I did search for SiC driven DRSSTCs on google but didn't find anything. looks like I should have checked here too.

Thanks for the tips on the resistor, and routing changes, I will implement them next chance I get. Any rough guidelines on the gate trace width and routing, I would be happy to receive some MS paint squiggles?
Also what's thoughts on adding an earthed shield pour to layer A under the GDTs? should help combat noise.

2oz boards are £45 but 3oz plus take it to the £100 territory. so board will be 2oz but not more, unless there were clear deficiencies.

honestly most of my prior projects have used IC based gate drives. for niceties such as automatic shoot through protection some even with internal bootstrapping.
most of my past experience is pick an IC, find a reference design, and you are mostly there. I wanted to used GDTs as its a bit of tech I haven't used before.

Once I have a working coil and I understand some more of the base principles, I like the idea of trying a SiC based "ACW" DRSSTC (almost continues wave.) Basically have a standard DRSSTC optimized to run with pulse skipping / freewheeling that runs a near constant peak current limit but no interruption with the goal of better fidelity audio modulation but no primary impedance issues a CW SSTC faces.

Anyways for now I will focus on getting this basic one working first 

[flyingperson23]

As Anders mentioned, SiC conduction loses are proportional to I^2 while IGBT is proportional to I. Even if you used expensive high end SiC fets with say 10mohm on resistance, the conduction losses would be the same as a standard $5 IGBT at 150A and twice the IGBT loss at a more reasonable current of 300A. This is why you don't see any SiC drsstcs, the main benefit of SiC is if you want to do high frequency or hard switching at lower currents, like in a qcw coil.

[Beggernator.]

That is very true. I forgot to factor in hard vs soft switching. As my prior projects have been relatively low voltage motor drives where everything was always hard switching.

I suppose the exception would be very high frequency tabletop coils where most losses would be switching related and not from condition.

I am still interested in a "ACW" DRSSTC coil in future but I guess as you point out IGBTs are still king here.


Changes were made to the bottom half of the bridge, with the top remaining in V1 trim so there is good comparison.

I added a shield pour under the GDT connected to earth. reworked the C and K IGBT connections. made sure that min clearance between all nets was > 2mm, apart from the IGBT legs, which will get a few coats of PU varnish once tested successfully at 100v or so.
Via spam is to help thermal management and current sharing into the main copper pour, will get varnished too so not worried about arcing.

Thanks again for the feedback.

(Edit to consolidate messages and insert picture)

[Anders Mikkelsen]

Ditch the shielding plane under the GDT, as it is now it will have significant capacitance to the switching node, and combined with the inductance of the long grounding wire you have a distributed element resonator that can cause noise coupling. Other than that it looks good to me.
 
For the GDTs, make sure you get the -B suffix part as specified in your schematic, the -A variant has a different pinout, I'm sure this is already considered but it's easy to miss.

[Beggernator.]

Thanks Anders. I had considered the capacitive coupling but hadn't thought about the inductance of the GND track making it susceptible to oscillation.

I have a Bom save on mouser and it is definitely the -B suffix the -A would have complicated routing so I didn't use.

Next I am looking at driver board options.

The UD3> driver looks properly swish, but it almost feels like cheating. Part of my wish to start with a smaller coil at 2 - 3KW was to understand the fundamentals and and having an FPGA based controller run some scans and spit some data at me would skip some of that learning.

the UD2.7c looks pretty swish. nice features, makes it harder to detonate your inverter. As far as I can tell it doesn't have phase skip implemented.
UD2.9 from Dan on this forum seems to be similar to the 2.7c but with the inclusion of skip pulse. Unfortunately I cant seem to find much info on it, BOMs don't seem up to date ect.

So with this in mind does anyone have any suggestions for a good driver with skip pulse that that is still somewhat traditional analogue electronics (smd or THT is fine).

[flyingperson23]

Pretty much all of the information for the ud2.7 applies to the 2.9, the 2.9 just has a little bit of extra logic for pulse skipping. It is a bit old, but it's been used in lots of coils and is known to be reliable.

[Beggernator.]

Thanks, That makes sense, I will compare the schematics when I get a chance see what differences there are.

Are there any dew drivers that are still discrete logic instead of programable / software based?

[flyingperson23]

There have been a few designs floating around here but none are anywhere near as tested and reliable as the ud2.7/2.9. Afaik there are a few improvements a newer analog design could have but nothing really performance improving and there are no designs with the kind of documentation and proven reliability the 2.7 has.

[Anders Mikkelsen]

I'd encourage people to play with untested circuits, because if nobody does then they remain untested and there is no progress. I can understand that people are replicating ten year old circuits again and again as they are known to work and proven reliable, but doing more research will benefit everybody

[davekni]

So with this in mind does anyone have any suggestions for a good driver with skip pulse that that is still somewhat traditional analogue electronics (smd or THT is fine).

I'd encourage people to play with untested circuits, because if nobody does then they remain untested and there is no progress.

Agree with Anders, would be great to get more experience with updated drivers.  If you are willing to consider such, I'd recommend:
   https://highvoltageforum.net/index.php?topic=2054.msg15611#msg15611
For further updating, above could be modified with Anders' proposed phase lead option:
    https://highvoltageforum.net/index.php?topic=3097.msg22172#msg22172

BTW, IKZA50N65EH7 looks like a great part, at least based on data sheet.  For pushing such TO247 package parts, thermal conductivity to heatsink is important.  I'd suggest considering copper heat spreaders under FETs or even separate electrically-hot heat sinks as I typically use.  And as you probably already know, apply force above FET die rather than using screw hole alone.

[Anders Mikkelsen]

I would definitely second David's note on the thermal interface, especially if going for high power throughput. The thermal resistance of the interface, together with the power dissipation and heatsink temperature, determine what the package temperature is. The higher the case temperature is, the lower the pulse handling capability since you want to stay under the maximum junction temperature.

Sil-pads are among the worst of the practical solutions, typically adding 1 - 2 k/W of thermal resistance for a TO247 (assuming 160 mm^2 heat spreader area), so with 50 watts of dissipation the package will be 50 - 100 degrees above the heatsink temperature. Their only advantage is that they are easy to use.

Kapton foil can offer better performance. Plastics are generally not very thermally conductive, so this is achieved by making the interface very thin. I've used this in the past, but I had failures due to imperfections in the heatsink surface. Capacitance between the Collector and heatsink is also pretty high since the interface is so thin. These can be made to work well, but they can be very finicky to get reliable.

I had the best success with ceramic insulators, since they have good thermal conductivity and very high dielectric strength. Aluminium oxide is the most common and one of the cheapest ones, with pretty decent thermal performance. 1 mm thick Alumina will add around 0.25 k/W for a TO-247, and 0.6 mm will be even better at 0.15. For even better performance Aluminium Nitride can be used, but it's also significantly more expensive. With Nitride, the thermal resistance for a TO-247 interface can be below 0.05 k/W.

Another bottleneck is the thermal compound, especially when using ceramics. For best performance, you want a high thermal conductivity compound and an even layer as thin as possible bridging the gaps. It pays to make sure that the heatsink and ceramic insulator are flat, to not have any unneccessary gaps to bridge with the paste, as even good pastes have ten times higher thermal resistivity than Alumina. One thing to keep in mind is that the highest thermal performance compounds tend to be very viscous, having a consistency closer to chewing gum, so the highest conductivity compounds will not always necessarily be the best, unless a very even layer is applied initially and a lot of pressure is used. I also recommend avoiding thermal compounds with metallic filler, as they can cause shorts between transistor pins or between the package and heatsink unless carefully applied and cleaned after installation.

[Beggernator.]

Thanks for the great replys everyone.
Honestly I am really liking the look of the ud2.9 but I'm not a great fan of some of the existing board layouts.
I will have a read through the other threads and who knows we might end up with a UD2.99

From the thermal management side of things I was actually already had a setup in mind. I have some 5mm thick 20mm wide copper busbar left over from a battery project so each igbt was going to get a chunk of that direct to it for added thermal inertia between bursts and then that was going to go to a shared heatsink via electrical isolation. Initially I plan for the heatsink to be air cooled but potentially might get swapped for a water block if I fancy turning it up to 11.

The PCB design I have shown above has holes for through bolting to a heatsink allowing the igbts and thermal stack to be under compression for better thermal conductivity. There will be a spreader plate on the back side so the force isn't just flexing the PCB. (Probably 3d printed, glass reinforced asa melts higher then the igbts can take so won't be the first point of failure).

On the note of thermal interfaces I think it's worth mentioning PTM7950. Some of the PC modding community have tested this material on direct die silicon and claimed to reach near liquid metal thermal numbers with it. I tried some previously with limited success due to poor tolerances.  While it is non conductive I do not believe it has the dielectric strength to isolate the igbts on its own but as an interim between the igbts and the copper spreaders I feel it's worth a go.

Great detail on the isolation marterials anders. It's not something I have particularly looked into before. I guess price to performance 0.6 aluminium oxide is the best bet. I am quite used to viscous thermal pastes, the one I have in my graphics card at the moment needed preheating to 60 degrees to be properly workable.

Also lapping components probably helps the thermals. Will need to find someone with nice granite kitchen counters I can stick some wet and dry sandpaper too.

It's also good to hear others approving of the IKZA50N65EH7 igbts. My slight concern is the copack diode is only rated for 80A peak so pulse skip might give it a hard time. I suppose worst case I can add some diodes in or not run pulse skip. Time will tell I guess.
(Edit spelling)

[davekni]

The PCB design I have shown above has holes for through bolting to a heatsink allowing the igbts and thermal stack to be under compression for better thermal conductivity.

This places compressive force at IGBT screw holes rather than over IGBT die.  Any compliance in the thermal stack will compress most there and can even tilt part ever so slightly away at opposite end of IGBT packages, which significantly increases thermal resistance.  There are clips designed to use screw hole while applying force over die, such as:
    https://www.mouser.com/datasheet/2/2/Aavid_02292019_MAX08NG-1539505.pdf
However, these wouldn't work through your ECB.  Perhaps you could print a glass-filled 3D part that mounts with two of the screw holes on opposite sides of ECB with bumps over IGBT die to apply force there rather than at hole locations.

I have some 5mm thick 20mm wide copper busbar left over from a battery project so each igbt was going to get a chunk of that direct to it for added thermal inertia between bursts and then that was going to go to a shared heatsink via electrical isolation.

Sounds great.  Cut sections as long as will fit symmetrically under each IGBT without touching section of adjacent IGBT.

PTM7950

Thank you for mentioning above material.  Honeywell data sheet looks great, though I haven't found any corresponding mounting pressure required to achieve specified thermal impedance.

Also lapping components probably helps the thermals. Will need to find someone with nice granite kitchen counters I can stick some wet and dry sandpaper too.

Good thermal compound should handle minimal roughness of normal parts.  Perhaps lapping would help further.  This is the first time I've heard lapping suggested.

My slight concern is the copack diode is only rated for 80A peak so pulse skip might give it a hard time.

I see 200A diode peak rating on middle line of table 4 in spec you linked earlier.
Diode thermal capability is 25% lower and forward drop is ~25% higher than for IGBT, so diode couldn't handle as much average current.  But even with freewheeling, IGBTs are conducting more of the time than diodes presuming some power is being transferred to secondary (and to coil losses etc.).  I'd guess these parts will work quite well.

[Beggernator.]

Again thanks for the insight Dave.

At some point in the near future I hope to have some CAD that will be a better display of what I am proposing, I will do some quick and dirty sketches to hopefully convey my intentions at the end of this.

I see 200A diode peak rating on middle line of table 4 in spec you linked earlier.

Great spot I totally missed that, been looking at hundreds of power silicon datasheets so must have got wires crossed. good news too.

Thank you for mentioning above material.  Honeywell data sheet looks great, though I haven't found any corresponding mounting pressure required to achieve specified thermal impedance.

This is purely anecdotal evidence, but people have been using it inside high performance laptops where the die mounting pressure is achieved by incredibly thin bits of spring steel, must only be a few newtons or so spread across the whole die area. (To this end Lenovo have started using it from the factory on top end models, 70+ wats of dissipation from silicon only 10 to 12mm big)

When I tried it it was on my 3090 graphics card. I saw average temps drop compared to traditional high conductivity thermal greases while under the same mounting pressures, unfortunately the heat spreader on that graphics card has a noticeable dip near one corner so my hotspot temps went through the roof, swapping back to regular thermal grease sorted that out.

As I understand it the PTM material changes phase from a solid pad to viscous liquid at 40 to 60 degrees. due to this some bake in time is expected where elevated temps are required for it to properly conform to heat spreader surfaces.

Perhaps lapping would help further.  This is the first time I've heard lapping suggested.

Its a fairly old school PC overclocking practice, As you might have guessed from me keep talking about pc bits its where a lot of my early experience comes from.

People would lap the surfaces of the CPU heat spreader and the corresponding face on the heatsink to near mirror finishes on a glass plate, usually an old window pane or something. and usually a decent temp improvement could be seen from doing this. I remember having an old amd bulldozer CPU running near 240W of power after lapping at similar temperatures to what it reached at 200w of power un lapped. but there was a decent amount of concavity to the heat spreader on that chip.

Do I think this is really important for power electronics? probably not. I mostly thought about it for the copper busbars as they are very rough tolerance raw copper.

Any compliance in the thermal stack will compress most there and can even tilt part ever so slightly away at opposite end of IGBT packages, which significantly increases thermal resistance.

I had a neat trick in mind for that. at work we have some very thin polymer sheets. these are used in the stack up of pouch cell lithium batteries, it is essential that mounting pressure is equal across all parts of these cells so this polymer foam is designed to equalize mounting pressure across the face of the cell. I was going to place some of this foam between the IGBTs and the PCB so when the whole stack is compressed any imperfections will even out.

It is also worth mentioning I plan to assemble and install the thermal stack-up before soldering any of the IGBT leads into place to keep solder joint stress at a minimum.


(I forgot the insulators in pic 1)


bolts are blue, plastic reinforcement is green, rest is hopefully self explanatory, will be thermal grease applied at all interfaces.

The copper spreaders will likely get a small bolt through the IGBT hole but this will not continue to the aluminum as isolation needs maintained.

I will have some questions about secondary windings soon. I have the 82mm upvc form and 0.224mm class 2 enamel wire. i have planned dimensions written somewhere I just need to find them before asking opinions. 

Thanks
Tom

[Beggernator.]

I found the secondary information I had put together previously.

It started off innocently enough looking at Mads website for secondary coil design notes. (extract shown below)



In my head for a beginner coil keeping it on the smaller side is beneficial, I had a target of fully utilizing a 16A Commando socket so near 4kw. (yes I know it will be a fair stretch for the small IGBTs but be nice to to make a secondary I can grow into so to speak)

At the same time pushing as much power in as small a footprint as possible typically complicates things so I didn't want to make the most compact coil possible either.

Basically this was a very long way of saying as per the uploaded image of mads table I felt I wanted something in-between the sizes he mentioned.

Due to this I settled on 82MM outer diameter coliform and ordered a 500mm section of U-PVC pipe to serve this purpose. (actually ordered some solvent weld abs pipe but it ended up being U-PVC).

With regards to aspect ratio I notice many newer DRSSTCs are going for lower and lower aspect ratios. The original chart on Kaiser power electronics suggests 1:4 or 1:5 width to Hight for secondary's in this size range. but looking at modern coils like franzolis alpha DR and alpha DRmk2 or indeed Madzes own mk4, suggest an aspect ratio closer to 1;3.2 to 1;3.8 being ideal.

Personally I prefer the look of a taller skinner coil with large toroid so I was planning to target an aspect ratio of 1:4 being on the high side of modern coils but still well within the range of recommendations.

Next was wire choice. again I effectively made a educated guess. I ended up ordering some 0.224mm OD wire, (This is just the copper size, for winding calculations I will need to measure the total OD including insulation.)

I chose this size as it seemed to be about a happy medium between the 70mm and 100mm coil entries in the secondary coil size table.

Wire choice was a little limited by what I could find in the UK in the quantity required. But I managed to get lucky and scored a kg of wire in class 2 trim with 200degree heat resistance.
Interestingly upon inspection of the datasheet (which I did not read when purchasing  ) the base coat of insulation is polyester, with the second (topcoat) being polyamide. not sure how common this is but seems pretty cool.

For varnish I have 500ml of PU based transformer enamel, its rated for air dry, and 70kv in 1mm thickness. dip or brush application, I see no issue with this other then the suspicion 500ml might not be enough.

Plan is to wet wind it in one sitting, I have a willing sacrifice volunteer, so I wont need to grow an extra hand to apply varnish and wind. I am sure with the double insulation wire construction and thin viscosity transformer enamel wet winding isn't needed but I want to move this coil around and I think the mechanical strength of a wet wind will help it travel better.

TLDR:
82mm OD U-PVC Form
Target 1:4 aspect ratio so 328 mm long winding, will leave some space below to adjust Hight above primary and thus coupling coefficient.
0.224mm wire. at a guess additional 0.02mm for insulation, so call it 0.245mm, 328 / 0.245= 1338 turns. possibly a bit low for a coil of this size?
Upping the aspect ratio to 1:4.5 gives a height of 370 and an expected turn count of 1510 on the high side of the sweet-spot now. but I would rather be on the high side because winding inconsistency's will be more likely to reduce turn count, not increase it.
As an edit measured the wire with some pretty crap digital calipers, seemed to be around 0.23-0.24 so perhaps final dimensions of 82mm x 350mm would equate to around 1480 turns and be closer to my desired 1:4 aspect ratio.

I suppose it also begs the question is it better to have a desired turn count or height in mind for winding a secondary coil?

not sure, and any input is welcome. its my first time and all that  .

Cheers
Tom.

(edit for additional information)

[davekni]

bolts are blue, plastic reinforcement is green, rest is hopefully self explanatory, will be thermal grease applied at all interfaces.

Nice "sketches"!  Far better than anything I can create in the way of mechanical drawing.  If I'm now understanding correctly, force is applied to IGBT bodies through four corner mounting screws clamping shown sandwich.  I'm not remotely close to an expert in mechanical design.  Looks like force uniformity depends on amount of bending of your green stiffener layer compared to compliance of added plastic foam layer.  Maximum force would be on outer corner of each IGBT.  May be fine.  I'm just used to designs that apply force to center of each package, what I see in commercial power supplies and use myself.
BTW, have you considered how easy or hard it would be to replace an IGBT or two in case of failure?

I had a target of fully utilizing a 16A Commando socket so near 4kw. (yes I know it will be a fair stretch for the small IGBTs but be nice to to make a secondary I can grow into so to speak)

I'd guess 4kW is within reason if your design is towards high impedance and long on-times (high duty cycle).  From what I see on this forum, that's the way to get best performance for given H-bridge parts.  My DRSSTC is at opposite end, low impedance and resulting shorter duration impulsive arcs.  Loud, but not good for MIDI sound quality with more than one or two notes at a time.  Built before finding this forum or any other such information sources.

Interestingly upon inspection of the datasheet (which I did not read when purchasing  ) the base coat of insulation is polyester, with the second (topcoat) being polyamide.

Perhaps reasonably common.  Last week I'd purchased a spool of 19AWG wire with same insulation layers for my planned CW TC plasma speaker.

I suppose it also begs the question is it better to have a desired turn count or height in mind for winding a secondary coil?

JavaTC results for your coil design (including top load and primary) would allow for better decision making.  That will show resonant frequencies, coupling, and impedances.  For longer on-times, I'd stay at or above 50k ohm secondary impedance even if that increases aspect ratio.  Lower aspect ratio is good for high coupling without racing sparks, but not worth going below 50k for best chance of achieving 4kW.  (My DRSSTC aspect ratio is almost 7:1, 110cm/16cm.  Not what I'd design now, but still works well, especially with my redesigned primary coil that minimizes racing sparks by minimizing induced secondary volts/turn.)

Very well thought-out project!  I'm guessing you'll have great performance when completed.

[Beggernator.]

Thanks Dave.
They were quick and dirty, a screenshot of the 3d board from Kicad and some boxes drawn in Powerpoint. I am quite lucky in that I can visualize things in 3d in my head, but turning that into words sometimes eludes me.

Maximum force would be on outer corner of each IGBT.  May be fine.  I'm just used to designs that apply force to center of each package,

You are quite correct, as I mentioned most of my experience is computer based and have had multiple KG of mounting pressure in a similar setup to this without noticeable issue. it is possible this will be different but wont know until it is tried.

I have however thought of potential corrections if needed, I can print a different green backing plates that has gaps at the corners where it mounts and only touches the PCB behind where the IGBTs sit. giving that same effect as applying force to only the center of the package.

As for swapping of components it will be a pain. will need to remove the whole thermal stack up to swap the parts. not ideal by any means. but nothing is actually getting glued so its all doable, just a bit fiddley.

I'd guess 4kW is within reason if your design is towards high impedance and long on-times

yeah I had seen about a preference for high impedance on musical coils. I guess that makes sense thinking about it. with these H7 IGBTs low conduction losses they probably suit longer on time higher impendence setups (lower peak but higher average current).





(and no I am not sure why the sample view window is that scuffed. I will try downloading and running it later. instead of just the web view one.)

This is current java TC I have for the coil. It all feels very chicken and egg and a little bit confusing.
My toroid will be 3D printed and foil covered. so I can kind of chose what I want within reason, and I am choosing dimensions for my secondary coil. so feels a bit like too many options.

I have a kind of mini goal that I want the frequency under 200KHz just seems like it would be kinder to the IGBTs.

its looking to be around 45Kohm so a hair under that magic 50Kohm you and others have mentioned.

Looking at JAVATC it would also appear I fucked up on ordering a tank cap. I got a pair of dawncap 0.44uf 6000v ones, waaay too much tank cap by like an order of magnitude, oh well, they were cheap on eBay and can save them for a future coil.

Thanks again and Looking forward to your input.