Help with Class E plasma speaker Eastern Voltage Research

[FonziDaytona]

Hi all,

I need some help with my Class E Tesla coil/plasma speaker from Eastern Voltage Research. I currently have it playing audio over the discharge, but it murders very expensive MOSFETs. Technical support hasn’t been very helpful. They said that out of 10,000 customers, I’m the only one having trouble.

If I start the audio before starting up the coil, I kill both MOSFETs instantly. If I raise the primary 1/4” to increase coupling (recommended by EVR instructions) I kill both MOSFETs instantly.

Is this pretty much expected for these type things or is there something really wonky that I could check? I feel like if I look at it wrong, I’ll blow another $20. I’ve wasted $100 on MOSFETs now. I have video of it working as proof…

Edit: Here’s a video of it working just to prove that I did well enough for that. Problem is that it just eats up MOSFETs. My only idea left is that Dan’s recommended alternative to their original MOSFETs (they don’t have anymore for sale) just isn’t a good enough alternative. I tried irfp460 MOSFETs and blew them all within seconds which leads me to believe that is the case here. Failure is always at startup.

https://youtube.com/shorts/9PG3QSf6wqE?feature=share

-Matt

[FonziDaytona]

Here are some scope shots of operation. First is the gate to source signal pre Tesla coil operation. The second is the gate to source signal while operating. Last is the drain to source while operating.
No audio on these runs, back to trying to tune.

The first and last look somewhat okay to me, but the second looks pretty awful. Down to just one more set of MOSFETs.





-Matt

[Mads Barnkob]

If EVR is serious about your unit being a 1:10000 customer fault, they should issue you a new unit as this would then be malfunctioning, in their own words?

What is this 4 MHz signal written on screen? Surely not the drive frequency of the unit?

[davekni]

Thank you for posting scope pictures.  That makes it possible to offer some assistance.

What is this 4 MHz signal written on screen? Surely not the drive frequency of the unit?

It would appear to be the drive frequency.  I was not aware of any fixed-frequency class-E coils other than my two ISM-frequency ones.  Must be dissipating significant power in gate drivers.

Are there schematics available for this board, or at least the output section?  That would further help.  Are the two FETs in parallel for higher drive current?  Most class-E coil schematics I've seen use single FETs.  Are there any primary resonant capacitors beyond FET drain-source capacitance?

The issue appears to be too-low a primary resonant frequency.  Gate drive is low for roughly 100ns of every 250ns period.  Drain (primary coil) waveform should return to zero within that 100ns due to resonance.  In your case, the FET(s) are turning on with high drain voltage (~250V).  That causes excess FET power dissipation, so is the likely cause of your FETs failing.  The messy gate waveform is due to this rapid fall in drain voltage.

Low primary frequency must be either high capacitance and/or high inductance.  For inductance, the first step should be twisting or taping together primary lead wires.  Space between primary leads adds non-useful inductance.  You may need to remove a turn or two of your primary winding too.  For capacitance, if there is an additional resonant capacitor, perhaps it's value is incorrect (too high).  Or, FETs might be wrong part number (or counterfeit) and have high drain capacitance.  You may even have better luck running with only one FET (presuming the two are in parallel) to reduce capacitance, if there is no additional resonant capacitor to remove or reduce.

Whatever changes you make, scope the drain waveform as you did here during as-short-as-possible run times to reduce frying chances.  See if you can get drain waveform to fall back to zero or almost-zero volts before gate turns back on.  Also make sure drain voltage doesn't get too high.  Raising frequency will also raise peak drain voltage.

One more thought:  Perhaps this board has the wrong crystal oscillator frequency.  If it is supposed to run at 2MHz or 2.5MHz rather than 4MHz, then your primary coil would be about correct as it is.

[FonziDaytona]

Hi everyone, thanks for the replies.

EVR is standing by “user error” despite me showing its playing audio. After calling their product “fickle” (definitely seems so) and letting them know that I’d look elsewhere for help, they stopped communication. I’ve asked them many times to explain what to check and why I could be playing audio but also wrecking MOSFETS. Oh well. 

Anyways, the MOSFETs came from digikey so should be legit.

I’ll try twisting the primary leads together to see if it improves the waveforms at all.

I have a schematic, but will have to see if EVR has anything in writing stating that it can be shared, or not. I suspect that I’m not their favorite customer right now.

4Mhz is correct. It comes straight from a crystal oscillator.

One MOSFET is used for audio modulation while the other does the switching.

I’ll look over the schematic to see if I can scan the power section.

Edit: Killed my last set of MOSFETs. Twisting the primary leads didn’t help. I’m so sick of this thing and EVR.

-Matt

[FonziDaytona]

Here’s a picture of the power section.


-Matt

[davekni]

One MOSFET is used for audio modulation while the other does the switching.

Oh, that makes sense.  Thank you.

Here’s a picture of the power section.

That helps.  Can you copy any more?  I see a bit of instructions for winding L101 (primary coil), but they are cut off.  Schematic suggests a tapped primary coil (upper connection not at top of winding).  Are there any instructions about trying fewer turns (tapping at perhaps 4 turns instead of all 5)?  Any instructions about stretching the primary coil vertically (like a stretched spring), which would reduce inductance?  Somehow either capacitance is high (C10 is more than 560pF) or inductance is high (L1 is more than it's nominal 7uH or your coil windings are higher than expected inductance).

If you decide to experiment any further, can you power just this final stage from an external lower-voltage supply?  Then only one FET fries if something goes wrong.  You could tune your coil at lower power with less risk of frying even the one FET.  FET drain capacitance will be a bit higher at low voltage, but you could at least get close that way.

BTW, I've purchased only one item from Eastern Voltage Research, a MIDI to interrupter board.  It was easy to build and worked, but had several issues.  Got completely confused with more than two notes simultaneously.  Even single notes were mis-tuned, off by as much as 2% in frequency.  I offered to fix their processor code if they shared it.  They said the code came from someone else and they didn't have source.  I designed and built my own converter after that.

[FonziDaytona]

Here is the specs for the primary.

Also, I checked c10. I was given a 500pF (501J marking) to install when the instructions say it should be 560pF. Could this be the culprit?

Edit: Looks like I have one MOSFET left. Any harm in just bypassing the modulation MOSFET and use the on board PSU fed by a variac to drive the single MOSFET? The power stage has its own 140VA 70v power transformer.



-Matt

[davekni]

Here is the specs for the primary.

Thank you.  Makes it clear that the windings should not be spread out.  Looks like your primary winding is per directions.  (I recommend keeping the leads twisted.  Twisting or otherwise pairing leads close together reduces inductance slightly.  Didn't expect that alone to be sufficient to fix the problem.)

Also, I checked c10. I was given a 500pF (501J marking) to install when the instructions say it should be 560pF. Could this be the culprit?

No.  Resonant frequency is too low, so only a higher-value capacitor would be a possible cause.

BTW, did the kit include thermal pads (electrically insulating and thermally conductive) to go between FETs and heat sinks?  After all the FET replacements, it might be good to replace pads too.  If they get contaminated or warped or whatever, thermal contact resistance may increase, contributing to FET overheating.  Pads will also have some effect on drain capacitance, but not likely a large factor compared to total capacitance.

Edit: Looks like I have one MOSFET left. Any harm in just bypassing the modulation MOSFET and use the on board PSU fed by a variac to drive the single MOSFET? The power stage has its own 140VA 70v power transformer.

Sounds like a great idea.  However, I cannot say anything confidently without access to complete schematics.

If you are interested in learning more as you experiment, I'd suggest putting the coil geometry into JavaTC.  That will predict (fairly accurately) inductances, coupling factor, and resonant frequencies.  Useful by itself or for defining parameters to use in a simulation (LTSpice or other).

[FonziDaytona]

Hi David, thanks for all of the help!

Here is the rest of the modulation section. The only concern I have is the MOSFET bias adjust section. If I remove the blown MOSFET and jumper Drain to Source, I should have everything bi-passed safely, no?

I guess this could help rule out the high side modulation as being the culprit, no?

Also yes, it did come with thermal pads. I have replaced them though.

-Matt

[davekni]

Hi David, thanks for all of the help!

You're welcome.  Hope it ends up with a functioning circuit.

Here is the rest of the modulation section. The only concern I have is the MOSFET bias adjust section. If I remove the blown MOSFET and jumper Drain to Source, I should have everything bi-passed safely, no?

I don't see any issues.  Bias circuit is fine without load.  Just make sure not to feed more than ~80% voltage in to Drain_Supply+.  When operating normally, the average output of this amplifier stage will be something around 60-80% of input depending on bias adjustment.  Run all initial tests at much lower voltage, perhaps 20%.

I guess this could help rule out the high side modulation as being the culprit, no?

Yes, it is possible (but unlikely) that this modulation FET is frying first, then frying the class-E FET due to excess average voltage.  The modulation FET is running in linear mode, not full on or full off.  Most modern FETs are designed for switching mode.  Even though data sheet lists SOA curves that indicate it should be fine here in linear mode, it may not have been characterized fully that way.  Hot-spot failures are common in linear mode - same issue as with BJTs.

Also yes, it did come with thermal pads. I have replaced them though.

Good.  I presume your replacements are of reasonable quality and performance.

Unless you have other high-frequency/voltage caps around to replace the 500pF one, I'd start experimenting by reducing primary coil to 4 turns and perhaps spreading the turns vertically, at least enough to match height of existing 5-turn coil.

Good luck!  Please keep us posted on progress.

[FonziDaytona]

Bypassed and powered up to 25% via the variac. Got some 1/2” arcs and took some scope shots. Yellow is the drain voltage and magenta is the gate voltage. Primary is still 5 turns and twisted together.

Should I try upping the voltage next, or adjusting the primary?

-Matt

[davekni]

Bypassed and powered up to 25% via the variac. Got some 1/2” arcs and took some scope shots. Yellow is the drain voltage and magenta is the gate voltage. Primary is still 5 turns and twisted together.

Should I try upping the voltage next, or adjusting the primary?

As expected, the primary is farther out-of-tune at low voltage because FET drain capacitance is higher.  FET turns on before drain voltage even reaches its peak. Should be after drain voltage returns to (at least almost) zero volts.  Don't increase input voltage until either inductance or capacitance is reduced.  Presumably inductance is easier to reduce, by fewer turns and somewhat by spreading out vertically the remaining turns.  Or by reducing diameter, but that risks arcing across secondary turns inside primary coil when coils are too close.

[FonziDaytona]

Reduced primary turns to 4 and changed things a bit.

[davekni]

Reduced primary turns to 4 and changed things a bit.

Not the response I would have expected.  Perhaps secondary resonance is affecting primary current more than I'd guessed.  Were you at the same input voltage?  Average drain voltage looks lower.  If Variac was set lower, drain capacitance is higher, which might explain observed waveforms at least somewhat.

In order to make useful suggestions, I'll need to make an LTSpice model, starting with JavaTC-estimated coil parameters.  If you have interest, I'd suggest starting down that path, at least with JavaTC.  It isn't difficult to use, though do pay attention to radius vs diameter parameters, and be sure to select inches or cm and use throughout.  I enjoy simulating and learning, so will likely get to this myself.  However, likely out 2+ weeks.  Family is coming tonight through next week.

Might be interesting to see waveforms with secondary removed, at this same low voltage.

[FonziDaytona]

It was at a lower voltage. Here is another run with the secondary removed and at the original voltage. Not much different. The higher the voltage the worse the gate-source signal looks. I’ll look into JavaTC.

[davekni]

It was at a lower voltage. Here is another run with the secondary removed and at the original voltage. Not much different. The higher the voltage the worse the gate-source signal looks. I’ll look into JavaTC.

Now looks more like I expected.  Notice that drain waveform is curving over towards flat before FET turns on now.  Previous 5-turn capture shows drain voltage still rising.  (I presume this capture is really 50V/div for drain, even though scope displays 5V/div.)

Gate (and drain) 40MHz ring is due to parasitic inductance of FET leads and ECB traces.  Will be much less once tuned so that drain voltage is low during FET turn-on.

Does your 4-turn coil have additional lead length (same wire as 5-turn coil)?  Is the entire lead length paired (twisted or taped tightly together)?  Are the 4 turns tightly spaced, or spread out a little to match height of 5-turn coil?  If it isn't hard, a picture or two of each setup when you post would make interpreting easier.

If you have enough wire around, you could make a separate 3-turn (spread to 5-turn height) coil with short leads.  That would make it easier to change back and forth.

What is below the sheet of plastic on which the coil sits?  Is there anything conductive within a couple inches (ie. metal table or wood table with screws or ...)?

[FonziDaytona]

Yup, I see the difference.

The other probe was set on the the x10 scale on the probe.

There’s a wooden table under the piece of plexiglass, probably some hardware in the vicinity.

The primary wire was shortened as part of the reduction to 4 turns. Here is a picture of it, wrapped in kapton tape.

I’m out of this particular wire (14 awg silver coated copper with PTFE). EVR was really anal about me using this and only this…Originally, when I had the most success, I was using 15awg magnet wire and high coupling. I’ll wind a three turn out of that other wire to see what happens.



-Matt

[davekni]

The primary wire was shortened as part of the reduction to 4 turns. Here is a picture of it, wrapped in kapton tape.

If I had to guess, 4 turns spaced significantly farther (perhaps one wire diameter or a bit more between turns) might be about right at higher voltage.  Still, 3-turn version will be helpful for low-voltage tests, to compensate for higher drain capacitance at low voltage.  Definitely space the three turns, at least to match the height of 5-turn coil.

Simulations will make this easier.  Should be able to roughly predict effect of drain voltage changes, so know what optimum inductance at operating voltage will look like at low voltage.  (What is the normal voltage at DRAIN_SUPPLY_+ input?)

[FonziDaytona]

I spaced out the turns the best I could, PTFE is some slippery stuff! It’s more of a very loosely wound primary than evenly spaced. Here is a scope shot. I see that every other cycle, the amplitude is much lower?

Normal operating voltage is 85v when using audio. When tuning (adjusting the secondary top lead length) it’s supposed to be set to 100v.

[davekni]

I spaced out the turns the best I could, PTFE is some slippery stuff! It’s more of a very loosely wound primary than evenly spaced. Here is a scope shot. I see that every other cycle, the amplitude is much lower?

Normal operating voltage is 85v when using audio. When tuning (adjusting the secondary top lead length) it’s supposed to be set to 100v.

Yes, PTFE is slippery.  Great for low loss at high frequency, though that isn't a significant concern for primary at 4MHz.  Would matter a bit more for secondary's higher voltage, but still not that significant at 4MHz for a single-layer coil.

It's that every-other-cycle behavior that needs JaveTC and simulation to understand.  Coil is locking to 2MHz sub-harmonic.  Since you are increasing primary frequency (lowering primary inductance), I didn't expect such behavior.  Must be jumping from upper pole to lower pole.  JavaTC will help understand where the two poles are, and simulation should show what may cause running at one or the other.  Goal is to get 4MHz operation with drain waveform curving all the way back to zero volts within each 100ns gate low period.  That is the way class-E is supposed to work, with no turn-on power loss in the FET.

If you want to see a drain waveform, look at the green traces in this post of my 13.56MHz class-E coil.  Gate drive is sine-wave (resonant), black trace.
https://highvoltageforum.net/index.php?topic=1638.msg12549#msg12549
Of course, there are parasitic inductance rings here too, even more pronounced at this higher frequency, though lower amplitude since running properly in class-E mode.

[FonziDaytona]

One thing that came to mind that I forgot about…L1, the 25T choke, is still part of the circuit. Should it be bypassed for these low voltage, no audio, tests?

-Matt

[davekni]

One thing that came to mind that I forgot about…L1, the 25T choke, is still part of the circuit. Should it be bypassed for these low voltage, no audio, tests?

No, bypassing 25T choke would present a high-frequency short to the drain, preventing operation and likely frying the FET.
Does remind me of a question:  Did the core come from EVR?  Can you tell the coating colors?  Would be worth verifying the core is the correct type and size.  (Unless you have a meter to read its inductance.)  If the core is incorrect, that could be the cause of issues.

[FonziDaytona]

The core came from EVR, it is red and looks like it matches up with Micrometals example.

I have been playing around and noticed a few things. One, if I keep the windings tight, and stay under 20%, the frequency stays steady at 4Mhz. If I add a little spacing to the primary it switches back and forth (2Mhz and 4Mhz). I have been playing around with JavaTC and see that the secondary resonant frequency is around 1.7Mhz…Shouldn’t it have been speced to give closer to 4Mhz? Also, I see that I got my secondary coil dimensions off a little. It should have been a winding length of 3.675” and I wound it to 3.375”. My eyes got the better of me…

Here’s a scope shot of the 3 winding primary, no spacing, at under 20% voltage input.

[davekni]

I have been playing around and noticed a few things. One, if I keep the windings tight, and stay under 20%, the frequency stays steady at 4Mhz. If I add a little spacing to the primary it switches back and forth (2Mhz and 4Mhz).

4MHz is desirable, but the resulting sawtooth waveform is not desirable.

I have been playing around with JavaTC and see that the secondary resonant frequency is around 1.7Mhz…Shouldn’t it have been speced to give closer to 4Mhz?

1.7MHz might explain the 2MHz issue.  However, that seems too low for the coil geometry.  Wondering if some JavaTC entries don't correctly match your secondary.

lso, I see that I got my secondary coil dimensions off a little. It should have been a winding length of 3.675” and I wound it to 3.375”. My eyes got the better of me…

That would imply slightly fewer turns, which would raise frequency.  Not sure if this is enough to be significant or not.  I didn't see a turn count specified in EVR schematic.

Here’s a scope shot of the 3 winding primary, no spacing, at under 20% voltage input.

Not what I'd expected.  Just makes it clear that I'm not being too useful, not until I get a simulation running.

[FonziDaytona]

This looks (I think) like I nailed it! I had to lower the primary to below the secondary bottom winding, so very low coupling, right?

[Mads Barnkob]

This looks (I think) like I nailed it! I had to lower the primary to below the secondary bottom winding, so very low coupling, right?

You can calculate the coupling with JavaTC: http://www.classictesla.com/java/javatc/javatc.html

[FonziDaytona]

JavaTC is suggesting moving the primary even further down (can’t go any lower really). Judging by the scope shot, it looks a lot like Class E operation, no?

What should be my next step? Pump up the voltage (more power) or should I wind a 4-5 turn primary and keep the same coupling?

I’ve noticed that corona is now breaking out. No affect on waveforms unless I draw an arc.

-Matt

[davekni]

This looks (I think) like I nailed it! I had to lower the primary to below the secondary bottom winding, so very low coupling, right?

Yes, that is a great class-E waveform.

JavaTC is suggesting moving the primary even further down (can’t go any lower really). Judging by the scope shot, it looks a lot like Class E operation, no?

What should be my next step? Pump up the voltage (more power) or should I wind a 4-5 turn primary and keep the same coupling?

Is the scoped waveform with your 3-turn primary?

As you increase voltage, drain capacitance drops, so optimum at high voltage will likely be slightly back towards your previous "bad" waveforms.  Perhaps slowly raise voltage and slowly increase coupling to see if you can maintain class-E as voltage increases.

Edit:  Thinking through all the experiments so far, perhaps the 3.375" vs 3.657" secondary winding height difference is important.  Secondary appears to be fairly high Q.  With fixed 4MHz drive, a shift of ~5% in secondary resonant frequency may be the cause of issues.  May make sense to add the 8-10 turns needed to reach 3.675" before trying to tweak coupling and primary coil parameters more.

[FonziDaytona]

That’s with the 3 turn primary with close spacing. I turned it up to 50% and here’s the scope shot. Looks like the gate signal is deteriorating as I go up.

I’ll see if I can cleanly splice in some more wire and wind on some new windings. If that fails, I’ll rewind the coil with some Essex 22awg ultrashield that I have.

Edit: Wound on another 10 turns of wire. No change in waveforms.



-Matt

[davekni]

That’s with the 3 turn primary with close spacing. I turned it up to 50% and here’s the scope shot. Looks like the gate signal is deteriorating as I go up.

Looks like primary resonant frequency is still just slightly low.  Drain voltage is dropping to around 30-50V before gate turns on, not quite all the way to zero.  That is likely why gate waveform gets worse.  I'm not sure whether to worry about the gate waveform directly.  FET power dissipation is likely low enough at 30-50V turn on to avoid frying.  (Had been ~250V in your early plots.)

Edit: Wound on another 10 turns of wire. No change in waveforms.

Interesting.  Thank you for the update.  Perhaps I can figure out why eventually with simulation.

[FonziDaytona]

I added the little loop that EVR recommends and this allowed me to raise coupling before getting the distorted waveform. Here is the latest run at 50%. Looks a little improved over the last one.

[davekni]

I added the little loop that EVR recommends and this allowed me to raise coupling before getting the distorted waveform. Here is the latest run at 50%. Looks a little improved over the last one.

Yes, looks like clean class-E operation now!  Hope you manage to keep that up to 80V operating point.

[FonziDaytona]

Took it up to 85v in (the EVR recommended setting for audio modulation mode) and it still looks good to me. I zoomed in to show the switching points. Looks like MOSFET turns on when voltage is near zero volts.

If this looks pretty good, I’ll install the audio modulation MOSFETs (Q8) tomorrow to see if everything is still looking good before trying some audio again.

[davekni]

Took it up to 85v in (the EVR recommended setting for audio modulation mode) and it still looks good to me. I zoomed in to show the switching points. Looks like MOSFET turns on when voltage is near zero volts.

If this looks pretty good, I’ll install the audio modulation MOSFETs (Q8) tomorrow to see if everything is still looking good before trying some audio again.

Yes, looks good.  Zoomed-in capture is nice, showing detail better.  Are you getting a reasonable plasma flame with this configuration and 85V?

[FonziDaytona]

It’s putting out a very bushy 1” discharge. Turns out it’s only 49v, I have another 36v to go till I get to the EVR recommend level. I may have to increase primary turns as I’m blowing my 5A fuse. At least I’m not blowing MOSFETs anymore! I’ll report back tonight.

-Matt

[FonziDaytona]

Here is the Tesla coil running audio again. You’ll have to excuse the teddy bear and the glass of whiskey, they are for emotional support!

https://youtube.com/shorts/6GjAaDhPUnA?feature=share

[davekni]

I may have to increase primary turns as I’m blowing my 5A fuse.

Great to see it working.  Increasing primary turns may lead to the initial issue of drain voltage not returning to zero before gate voltage turns on.  Not certain.  Your added 10 secondary turns might have fixed that.  If added primary turns are necessary to prevent fuse blowing, you might need to reduce the value of the 500pF resonant capacitor to compensate.

[FonziDaytona]

Looks like the blown fuse was just a fluke. I replaced it and haven’t blown any others, so I’ll keep experimenting before making any changes. So far, so good! I appreciate all of the help and will follow up later on.

-Matt

[FonziDaytona]

Figured I was due for a follow up:

No more blown MOSFETs! I tried going up to 4 turns on the primary but couldn’t keep the Class E operation waveforms. I rewound with some much nicer Essex ultragrade wire (to the correct dimensions) and still found that 3 turn seems to be the sweet spot.

I can push up to over 3 amps and 70v rms but my 5A fuse blows after awhile. Audio is very nice and arcs/sparks/discharge is very bushy and hot.

It’s too bad that EVR didn’t suggest going down to a 3 turn primary, using a variac, and watching the waveforms. It would have saved me $100 in MOSFETs.

Thanks to you guys (Dave especially), I’m rocking and rolling! Yanni really shows off through this thing.

-Matt