induction heater issues

[aes92000]

Hello All, hope doing well.

i have finally got a new design working.
uploaded schematic and board design pics as well as others
top layer of board is the 'return' or ground of circuit. bottom is control signals
i have the system water cooled and plenty of air on mosfets and radiator
the coil leads are soldered to copper bars and then bolted to board

the issue i am having is when i use small coil ~105kH, the board starts to fry between contacts when under load. see picture.

the long coil measures ~49kH - works awesome with 1kg crucible
the short coil measures ~105kH

what do i need to check into to remedy the problem?
do i need to add more capacitors to tank?

[davekni]

I think there are a couple ways to look at the issue, just different descriptions of the same thing.  One is to say that the tubing leaving the circuit board is inductively-heating that spot on the circuit board.  Fields from the two tubes add between them.  The other way is to look at skin-effect.  Current will crowd to the edges where opposing flow is the closest.

Either way, I'd suggest adding copper sheet-metal extensions to the output terminals.  Ideally the copper planes overlap, with the necessary small bend and some insulator (ie. mica sheet) between them.  That why the close-approach area is spread out, so current is spread out.  Where the copper sheet-metal connects to the pipe, the sheet metal will be thick enough to block enough of the field.

A simpler and perhaps-good-enough solution is a single square of copper sheet metal soldered to the ground plane covering the output connections (covering the burn area and a ways around it).  That will block magnetic field, so protect the circuit board.

BTW, I'm calculating almost 240A RMS through the coil and caps at 105kHz and 54Vdc input, perhaps 230A after FET and inductor losses are included.  How thick is the circuit board copper?  Also, even with cooling, it will be interesting to see how those caps hold up with ~25A each continuously.  I use those caps in my DRSSTC MMC at well higher current, but low duty cycle.  The caps may survive fine with cooling.  My high-voltage low-duty-cycle abuse testing of these caps indicated good robustness.

[aes92000]

Thank you for response.

being a noob, i need to verify if i understand...

1."I'd suggest adding copper sheet-metal extensions to the output terminals."
see pic with rough drawing for what i understand to do
make bottom coppers to protrude out from board with tubing - see picture of my understanding


2. "A simpler and perhaps-good-enough solution is a single square of copper sheet metal soldered to the ground plane covering the output connections (covering the burn area and a ways around it).  That will block magnetic field, so protect the circuit board."
i am sure i follow exactly
are you saying to cover the burn area with copper plate to "beef up" that area - see picture of what i think you are saying


the pcb is standard double sided 1oz. copper

the capacitors work very well with the longer coil.
i have run for almost 3 hours melting copper and brass for ingots without any issues with any of the components.
i can melt 10oz copper, from power on to melted, no warmup, right at 6 minutes

[petespaco]

I see from your recent pictures that there is (or WAS) copper in the area that is burned.
In my opinion,  the copper is the CAUSE of the heating. Get if off of that area and I think your problem will disappear.

Do you see a big increase  in idle current when you use the short work coil?
Check your gate signals with a 'scope with both coils. You may find that your coil with only a couple of turns has too little inductance which may cause the Mosfets to turn on too slowly.  I would try to keep the frequency below about 80 kHz.

Pete Stanaitis
---------------

[davekni]

For your first image, yes, although I was suggesting that the added copper planes overlap in the center (with insulation between the two).

For the second image, yes, although I'd extend the copper piece down past the edge of the circuit board too.

In general, 230A RMS is high for 1oz copper even with the planes.  May be OK given your high air flow.  The idea of added copper is both to beef-up that area, but also to spread out the current flow, keeping it from concentrating in that small area.  That is the point of having the two planes overlap in the first suggestion.  Just beefing it up may be sufficient without overlap.

As Pete pointed out, at 105kHz the FET turn-on time will cause increased FET power dissipation.  If they are still cool enough, it may be fine.  Scoping gate and drain waveforms for one of the FETs will indicate how much switching delay (gate charging time) there is.  There are a couple circuit alternatives to reduce switching delays in my Jacob's ladder thread:
https://highvoltageforum.net/index.php?topic=831.msg5491#msg5491
Reducing the value of the 470-ohm gate resistors will also speed-up switching, but require even higher-power resistors.