Fixed frequency HFSSTC using gate drivers

[Lucasww]

I was looking at this thread and this page:
https://highvoltageforum.net/index.php?topic=2741.0
http://uzzors2k.com/index.php?page=4MHzclassE1
And was wondering if it would be possible to build a HFSSTC with frequency high enough to get a flame (8MHz?) using direct gate drive from a gate driver chip. It seems like the NCP81074A was able to drive a mosfet CW at 4MHz with no issues, so I feel like it should definitely be possible to drive a SiC FET at 8MHz with two of those chips in parallel.

I have some UF3C120150K4S SiC cascodes (18A 1.2kV) which lists a gate charge and input capacitance of 26nC and 738pF.
NCP81074A lists a maximum DC current of 0.6A. I ran some simulations in LTspice and got RMS current of ~0.27A for each chip with this setup:

Power dissipation is around 0.2W in each gate resistor

So is there any reason this wouldn't work? Provided I use good PCB layout it seems like it should be fine.

[davekni]

It seems like the NCP81074A was able to drive a mosfet CW at 4MHz with no issues, so I feel like it should definitely be possible to drive a SiC FET at 8MHz with two of those chips in parallel.

8MHz might be possible with good cooling of NCP81074A chip(s).  Paralleling two may have little advantage.  Most of the gate drive power will be dissipated in the UF3C120150K4S's internal 4.6ohms gate resistance.  Internal switching losses of NCP81074A likely cause much more NCP81074A heating than does external gate load, which is at the light end of NCP81074A's spec'ed use range.  If NCP81074A can handle 8MHz with no load, it probably can handle 8MHz with UF3C120150K4S load.  Use 2oz copper ECB, preferably thin (<=0.8mm rather than normal 1.6mm) with good thermal vias under NCP81074A's thermal pad to a large back-side ground plane.  Or use an aluminum-core ECB.

Notice that UF3C120150K4S is specified with -5Vgs to 12Vgs drive.  Even though in a 4-pin package, there is significant mutual inductance between the two source pins, creating internal Vgs spikes due to switching current.  I'd expect UF3C120150K4S switching power to be significantly higher if driven 0Vgs to 15Vgs.  I'd suggest adding a capacitor between NCP81074A output and UF3C120150K4S gate.  If constant duty cycle, a resistor divider can set average Vgs.  If not constant, a zener diode or shunt regulator circuit can set voltage across capacitor to 5V.

[Lucasww]

Internal switching losses of NCP81074A likely cause much more NCP81074A heating than does external gate load]Internal switching losses of NCP81074A likely cause much more NCP81074A heating than does external gate load

Ah, I hadn't even considered internal losses. That makes sense that it could cause more issues than the load. Is there any way to calculate these losses from information on the datasheet? I wasn't able to find anything in there. I've also been looking into other chips, supposedly UCC27524 and IXDD614 have been used for this. The IXDD in particular is interesting as they offer a TO-263-5 package which should be able to dissipate a lot of power without issues, however it's a bit slower and the larger size makes me question it's ability to operate at very high frequencies.
Also looking into this one:https://www.mouser.com/ProductDetail/Texas-Instruments/UCC27332QDGNRQ1?qs=ST9lo4GX8V0y1%252BN0eUmrLg%3D%3D
current is a bit lower but it has an exposed metal back and nicer pinout

[davekni]

Is there any way to calculate these losses from information on the datasheet?

Forgot to include this bit in my last response:  NCP81074A datasheet graphs of supply current are not self-consistent.  I spent some time trying to extrapolate to your conditions, but no way to do so with contradictory data within the existing graphs.  Resistors on each of the outputs (pull-up and pull-down outputs) will reduce shoot-through current and move much of the remaining shoot-through power dissipation to those resistors instead of driver chip.

Also looking into this one:https://www.mouser.com/ProductDetail/Texas-Instruments/UCC27332QDGNRQ1?qs=ST9lo4GX8V0y1%252BN0eUmrLg%3D%3D

I'd guess NCP81074A is still the best option.  No way to know for sure.  UCC27332 data sheet is quite clear even though graph goes to only 1MHz.  Extrapolating to 8MHz at 17V (for -5V to +12v), current would be 293mA no load at 8MHz, so 5W chip power dissipation.

[Lucasww]

I talked a bit with "UEC MITSUI Marine Diesel Engine" on youtube, who has used gate drivers for similar stuff, and he said that UCC27524 worked for him at 9MHz 15V without overheating. Since I know they will work, i'm going to use them instead.
Regarding the gate voltage, if it needs negative voltage I think I could do something like this

I would supply the chip with 15v and get around -3 to 12v on the gate.
That said, is negative gate voltage even necessary? The voltage will probably overshoot under zero volts anyway, but i'm not sure

Either way, I designed a PCB, heres a picture of the relevant part. I left the 2 source pins separate, and that's the only connection between the 2 grounds. Is that correct? Other than that, I think everything else is straightforward. C5 and C4 are 0.1u and 10u, gate resistors are 4.7R each.

[davekni]

he said that UCC27524 worked for him at 9MHz 15V without overheating.

Yes, looking at UCC27524 datasheet suggests it would be a good part.  Extrapolating to zero load looks like very close to zero internal switching current.

I would supply the chip with 15v and get around -3 to 12v on the gate.

Those three 1N4148 diodes will give you about -2V.  D4 is not needed.  BTW, a blue or white LED makes a good ~2.5 to 2.8Vf drop.

Either way, I designed a PCB, heres a picture of the relevant part. I left the 2 source pins separate, and that's the only connection between the 2 grounds. Is that correct?

Yes, looks good.

[Lucasww]

Well, my board arrived a few days ago and I put it together. The driver itself works great, the UCC27524s only get slightly warm to the touch after a minute of CW operation. I also built in interrupter support, I want to try interrupting it above audible frequency at some point to get lower power draw but hopefully a similar sized silent flame. But first I have to actually get the coil working of course.

Some pictures of the driver:



Gate signal looks as expected, starting at 0v slowly dropping down to around -2v. I wanted it to drop a little more but it probably doesn't matter.

And zoomed in. (Probe wasn't adjusted properly here, in reality the waveform is a bit more square. You can see it in later pics.)


And here's some class-E waveforms with no secondary coil, after a little bit of tuning. Primary capacitor is a 100pF C0G array.


All the following testing was done with 12-24V input into the class-E stage.

I start having problems when I add a secondary coil. I made a secondary with around 8.5MHz fres, dimensions around 2.3" diameter and 2" height. coupling between the secondary and primary has to kept under ~0.1 or it completely ruins the waveforms, like this:

I have absolutely no clue what is going on there.
If I keep the coupling low, I can get decent waveforms but the output is terrible. I get around quarter-inch arcs to a screwdriver but no corona with this setup:

Connecting the secondary to the node between the primary and capacitor works about the same if I flip primary phasing to match, but waveforms are worse. Flipping the primary phasing otherwise results in no output from the secondary whatsoever (not even able to light a fluorescent lamp). I also tried throwing together a secondary that resonates at twice the operating frequency, which seemed to work for Dave and Steve Ward, but it ended up with a slightly worse result.
Maybe I just need to fiddle with it more to get the tuning right, although i'm not sure. any tips about tuning a fixed frequency HFSSTC would be greatly appreciated, in particular I'd like to know about the effects of different secondary coils and coupling, but also whatever else is relevant.

Also, to Davekni, what input voltage did you use to start the flame on your fixed HFSSTC? that seems like it should be around the voltage I should expect to be able to get some plasma to start on mine.

[NyaaX_X]

These days I stuck..

But I still make some LTspice works for my 4MHz testing. But I think my power is on the wrong part.

You know the secondary resonator and plasma output can change the load of your class E stage. So you have to re-tune the Class E circuit.
From your waveform, I think you can add more the CTANK 100pF.

[davekni]

Maybe I just need to fiddle with it more to get the tuning right, although i'm not sure. any tips about tuning a fixed frequency HFSSTC would be greatly appreciated, in particular I'd like to know about the effects of different secondary coils and coupling, but also whatever else is relevant.

Secondary coil and tip capacitance form another resonant circuit.  Even if magnetic coupling to primary is zero, it is effectively coupled due to being in series with primary.  Some of the commercial kit pictures I've seen run this way with secondary placed far away from primary.
I found tuning to be non-intuitive.  Simulation (LTSpice in my case) is the best way I know for finding usable C and L values.

Also, to Davekni, what input voltage did you use to start the flame on your fixed HFSSTC? that seems like it should be around the voltage I should expect to be able to get some plasma to start on mine.

About 30Vdc input if I recall correctly.

Hope you find success with tuning!

[Lucasww]

Thanks for the tips. I messed with it some more and was able to get some small plasma at 20v with decent waveforms. I think I will try some different secondary coils later.
BTW, I did simulate the coil in LTspice before building it, but am having trouble getting the simulations to translate over to reality.

Dave, do you remember if primary phasing was important with your coil? I've never seen anyone mention it with regards to this type of coil but it seems like it should be important and in my experience it is. Also, is there a reason you used a secondary with the resonant frequency at twice the operating frequency in your HFSSTCs?

[davekni]

BTW, I did simulate the coil in LTspice before building it, but am having trouble getting the simulations to translate over to reality.

Yes, matching simulation to reality can be tricky, but it is worth the effort.  First make sure you have accurate values for inductances and coupling.  If wire length is significant, include that inductance too.
Next is to check FET model, especially for Cds.  If there isn't a good model available for that part, there is a free program called MOStool that you can use to make an LTSpice model from datasheet plots.
Finally, perhaps the hardest part is to estimate coil and breakout capacitance.  Perhaps JavaTC results are reasonable for that even though HFSSTC geometry is significantly different than most TCs.  If you have a signal generator and scope or other tool for measuring impedance vs frequency, measure self-resonant frequency of primary+secondary+breakout.  Tune breakout capacitance of model to match that frequency.

BTW, I did simulate the coil in LTspice before building it, but am having trouble getting the simulations to translate over to reality.

Since drive is fixed frequency, there is no feedback to have a phase question.  If you are referring to winding direction of primary and secondary matching, yes, that definitely matters.  Presuming secondary is on top of primary and connected in series with primary (bottom of secondary connected to top of primary) as is normal for HFSSTCs, winding directions should be the same.  Otherwise coupling factor is negative.

Also, is there a reason you used a secondary with the resonant frequency at twice the operating frequency in your HFSSTCs?

I didn't pay any attention to any possible resonant frequency of secondary coil by itself.  Did you calculate that my coil happens to be 2x operating frequency?  The possible issue I had initially was that the upper pole of the combined primary+secondary happened to be twice lower-pole (operating) frequency of primary+secondary.  Not sure if that was a real problem or not, but did create distorted waveforms.  I reduced breakout size to minimize that possible issue (increase upper pole frequency beyond 2x).

[Lucasww]

Yes, matching simulation to reality can be tricky, but it is worth the effort.  First make sure you have accurate values for inductances and coupling.  If wire length is significant, include that inductance too.
Next is to check FET model, especially for Cds.  If there isn't a good model available for that part, there is a free program called MOStool that you can use to make an LTSpice model from datasheet plots.
Finally, perhaps the hardest part is to estimate coil and breakout capacitance.

I put some time into simulating it more carefully, and settled on these values I was able to reach easily in JAVATC.
 [ You are not allowed to view attachments ]
 [ You are not allowed to view attachments ]
 [ You are not allowed to view attachments ]
(DS capacitance is pretty large here compared to other designs. I'm not sure why but this setup seems to like more capacitance there.)

I didn't pay any attention to any possible resonant frequency of secondary coil by itself.  Did you calculate that my coil happens to be 2x operating frequency?  The possible issue I had initially was that the upper pole of the combined primary+secondary happened to be twice lower-pole (operating) frequency of primary+secondary.  Not sure if that was a real problem or not, but did create distorted waveforms.  I reduced breakout size to minimize that possible issue (increase upper pole frequency beyond 2x).

I had calculated that, without the breakout capacitance, the resonant frequency was pretty close to twice the operating frequency in both cases. It's also close in Steve Ward's design so I figured there was a reason for that, but I guess it's just a coincidence. I designed the new secondary to have good performance with around 4 or 5 inches of flame, and it ended up with an unloaded resonant frequency around 14MHz.

Anyway, I built the new coil on a cardboard tube for lower dielectric loss (hopefully), and tweaked the primary coil to be around the right inductance.


And with a little bit of messing around, I was able to get a little bit of plasma at 20V
Heres a clip at 25V: https://photos.app.goo.gl/ngDT5sTMk1Ukh1p3A

And the waveforms are... well they're not terrible, but it definitely needs work.

(scope setup. One of my ground springs vanished so I'm using a piece of wire instead)

If I increase the primary inductance slightly, the waveforms look much nicer (basically like the better waveforms I showed in my last post), but there's significantly less output. Decreasing primary inductance even a tiny bit increases output but makes the waveforms significantly worse.

Unsure where to go from here as far as tuning, DS capacitance is already pretty high (300pF) but the voltage spikes still look too narrow. I think a good next step would be to build a more "permanent" base for everything so parts moving around doesn't affect the tuning.

[davekni]

If I increase the primary inductance slightly, the waveforms look much nicer (basically like the better waveforms I showed in my last post), but there's significantly less output. Decreasing primary inductance even a tiny bit increases output but makes the waveforms significantly worse.

Unsure where to go from here as far as tuning, DS capacitance is already pretty high (300pF) but the voltage spikes still look too narrow. I think a good next step would be to build a more "permanent" base for everything so parts moving around doesn't affect the tuning.

The high frequency ringing may be connection inductance to added DS capacitance.
Overall I'd guess your primary inductance is quite a bit too low and secondary somewhat too low.  Perhaps start with values from my 6.78MHz coil:
    https://highvoltageforum.net/index.php?topic=1589.msg12271#msg12271
Presuming top load and arc capacitance will be similar, probably most capacitances should also be similar.  That implies inductances should scale by roughly 1/freq^2, so ~2.8x inductance values of my 6.78MHz HFSSTC.  Of course, my values are not perfectly optimized.  No reason to go for exact scaling.  Just provides a starting point.  Looking back at my 13.58MHz coil compared to my previous 6.78MHz coil, capacitances are somewhat similar and inductances close to 4x, as would be predicted by this simple scaling theory.

[Lucasww]

Presuming top load and arc capacitance will be similar, probably most capacitances should also be similar. That implies inductances should scale by roughly 1/freq^2, so ~2.8x inductance values of my 6.78MHz HFSSTC.  Of course, my values are not perfectly optimized. No reason to go for exact scaling.  Just provides a starting point.  Looking back at my 13.58MHz coil compared to my previous 6.78MHz coil, capacitances are somewhat similar and inductances close to 4x, as would be predicted by this simple scaling theory.

I tried to keep fairly close to your values, but I assumed slightly lower arc capacitance because i'm not running it as high power as yours. It seems to be happy with 3.5uH primary with this secondary. Also my tank capacitor is a bit higher capacitance than yours, so lower primary inductance compensates for that.

I messed with it some more today, and I was suspecting that my waveforms were being messed up by noise a bit, so I added ferrites to my probe wires and it actually helped quite a bit. they're still a little noisy (especially the lower voltage gate signal) but it's passable for now. I was able to get the coil working decently well by very carefully adjusting the primary.





Video at 40V-55V: https://photos.app.goo.gl/XQ6JbKeMZgCbn27P6
(video takes a while to load sometimes, but it eventually works)

Annoyingly, the coil only works well this way within a range of around 15V. if I increase the voltage above 55V the arc suddenly shrinks, or disappears entirely.
Video: https://photos.app.goo.gl/rhJCcCx7pTXNhe3EA
I'm guessing this is the arc capacitance throwing it out of tune. I'm unsure how to get around it without just re-tuning the coil, but then I would have to start it at a higher voltage.

Also, this setup is extremely sensitive to the primary inductance. Moving the last turn up only a millimeter to barely decrease the inductance completely changes the performance. I think this might just be because my primary coil is very wide and short, so very small changes in it's length have relatively large changes in the inductance. I may try to make a new primary coil around the same diameter as the secondary coil. Also, the primary gets very hot, but it's already pretty thick. Would litz wire work better for this? if i'm already going to remake it then I might try that.
Another idea I had for tuning was to build the primary with it's inductance a bit too low, and have a small easily adjustable external inductor (just a few turns of like 1cm diameter) in series with the primary.

Of course this will slightly decrease the effective coupling between the primary and secondary, but I think the effect will be insignificant.

[davekni]

I messed with it some more today, and I was suspecting that my waveforms were being messed up by noise a bit, so I added ferrites to my probe wires and it actually helped quite a bit. they're still a little noisy (especially the lower voltage gate signal) but it's passable for now.

Yes, ferrites around probe coax cable helps in many situations.  Something I do frequently.

I'm guessing this is the arc capacitance throwing it out of tune. I'm unsure how to get around it without just re-tuning the coil, but then I would have to start it at a higher voltage.

Yes, I ran into that issue too.  I played with values in LTSpice trying to find where sensitivity to arc loading was minimum.  Still had some problems with tuning change.  Initial solution was to add NiZn high-frequency ferrite inside lower part of coil, then remove it as I increased voltage and therefore arc size.  That ferrite got quite hot.  My later solution is shown in my 13.58MHz HFSSTC thread.  FET is used to selectively remove a little capacitance from 75pF (in my case) primary resonant capacitor.  Your added primary inductance will likely work too.

Also, the primary gets very hot, but it's already pretty thick. Would litz wire work better for this?

The low impedance (low inductance and high capacitance) of your coil requires higher current, so may be increasing heating.  However, my primary does get very hot.  That's why I used bare wire and mica frame to hold it.  No kapton or other plastic to melt or burn.
Litz wire helps for perhaps 10kHz to 1MHz.  Above that it is hard make litz wire with fine enough strands to be useful.

[Lucasww]

Okay, i've been a pretty busy but I was able to get this working well a few days ago. I kept the same primary but added the tuning coil, which helps with adjusting the inductance a lot. I read from Steve Ward that adding extra capacitance hurts performance on HFSSTCs, but I tried adding a small "topload" anyway to help with detuning and that's what finally got it to work consistently.
The primary still gets very hot, I may try replacing it with 3.5mm copper tube. FET stays reasonably warm.
Flame at 80V input

I think i'm satisfied with this for now. I want to try a PLL version in the future, so you may see a thread about that soon.

[davekni]

I think i'm satisfied with this for now. I want to try a PLL version in the future, so you may see a thread about that soon.

Looks good.  Always nice to see projects finish working well.
I suspect a PLL version will avoid need to tune primary as arc grows.  Instead frequency will drop to match arc loading.  Same goes for any other way of tracking frequency such as the much more common self-oscillating versions.

[Lucasww]

I suspect a PLL version will avoid need to tune primary as arc grows.  Instead frequency will drop to match arc loading.  Same goes for any other way of tracking frequency such as the much more common self-oscillating versions.

Yes, that's what i'm hoping. I also want to try making the PLL version very low impedance so it can be operated in short pulses to get an effect like seen here:

(jmartis claims to be using PLL here, and it looks like they're using the "simplified" class E topology common for small tesla coils, since I don't see a choke anywhere.)

Maybe it could even be run QCW. I'm gonna try some LTspice simulations with PLL and see what I can do.