ZVS Driven Teslacoil

[Andrew321]

Hey all! I'm building a tesla coil driven by the primary of a ZVS driver. So far I have had some good success. I wound an 8" wide coil that's about 18" tall. I got parks out of it that were about 8" long at 60VDC input. Then I decide to play around and I wound a 10" wide coil that is about 20" tall. Its natural resonant frequency is about 107kHz and I tune up a ZVS to run at 106kHz. I put the secondary in the primary, hook it all up and suddenly its running at 120kHz. So i tune with the secondary installed. I drop the frequency down to about 95kHz and some small sparks are coming off. If I move my hand closer to the top of the coil the sparks get bigger. I assume because I'm adding capacitance to the circuit. But soon as i get to close or an arc hits me, the frequency jumps back up to about 120kHz.

I have heard a number of times this is a problem ZVS drivers have. Any tips on how to prevent it?

[davekni]

I have heard a number of times this is a problem ZVS drivers have. Any tips on how to prevent it?

Hopefully someone else will have some success experience to share.  This is more of an explanation.

As you appear to already know, ZVS-driven coils are inherently dual-resonant (DRSSTC).  That means they have upper and lower resonant frequencies (upper and lower poles) due to coupling.  That is true even if primary and secondary are tuned to the same frequency.  Normal DRSSTC drivers such as UD2.7 generally lock to the lower pole.  My experience with ZVS oscillators is that they generally lock to the upper pole.  Sometimes they get into complex modes that combine both poles (so not a sine wave output), especially when the upper frequency is a harmonic of the lower frequency.

My suggestion is to work with the upper pole.  Tune primary (ZVS) frequency to match the frequency of the secondary (w/o primary).  Then add secondary to primary and scope waveforms.  If getting a clean sine-wave at the upper frequency, then adjust other parameters (coupling, primary turn count, drive voltage) to improve performance.

[Andrew321]

Thanks for the insight Dave, always appreciated. I noticed this to be an issue with my 8" coil too. I just left it 1 turn primary wrapped tightly below the secondary and adjusted the capacitance until it worked. It was also way off from the initial resonance but worked about 90% of the time. I tuned it just below the working resonance so that approaching it didn't throw it out of whack and instead made the sparks bigger. But this new 10" coil is being difficult.

I was doing some reading on your plasma ball while googling this issue. Seems you added an RLC filter tuned to a target frequency to prevent locking to the upper pole. I was thinking I could give that a shot and just put it across the primary as you did. I will see what results that yields and then test tuning it to the upper frequency to see which works better. I'm also wondering if adding more turns to the primary would help. Maybe one turn is a bad idea?

 I love all this testing, just wish my garage heater worked!

[davekni]

I was doing some reading on your plasma ball while googling this issue. Seems you added an RLC filter tuned to a target frequency to prevent locking to the upper pole. I was thinking I could give that a shot and just put it across the primary as you did.

Coupling factor of my little ferrite-core transformers is much higher than for DRSSTC.  My filtering fix worked because the two frequencies were far apart.  With lower coupling and resulting closer frequencies, filtering would dissipate way too much power in the resistor of the filter.

I'm also wondering if adding more turns to the primary would help. Maybe one turn is a bad idea?

I'd guess more turns would be better.  Depends on your primary drive voltage, secondary inductance, etc.  Although this is a DRSSTC, you are running continuously (not interrupted), so I think SSTC designs are better reference points.  I see 4-6 turns typically on SSTC designs.  At 60Vdc into ZVS, voltage across primary will be 60V * PI / sqrt(2) = 133Vrms.  That's similar to many half-bridge designs, so perhaps 4 turns.  I'd spread out the turns to increase coupling.  Also, pair (twist or tape together) primary coil leads from coil to ZVS.  Otherwise lead inductance will reduce voltage reaching the actual primary coil.
If 1 or 2 turns ends up being best, using copper foil or other wide "wire" for primary winding will increase coupling.  1 or 2 turns can't be spread out effectively.

I love all this testing, just wish my garage heater worked!

Yes, experimenting is fun.  I also have an unheated garage.  Do have a heating pad on the chair at the desk where I solder etc., along with a heat-lamp just above the desk.

[Andrew321]

Your right, trying to filter out a frequency that close would not work at all. But never the less i have been testing! I started at 1 turn of wire just laid at the bottom of the secondary. I tested at 100kHz and 109Khz. Took notes then added 1 turn and adjusted the capacitance to get the same 2 frequencies. I did this up to 4 turns. I noted that with the primary tuned to both frequencies and then adding the secondary always dropped the frequency down about 12kHz. This is when output was largest and it decreased with additional turns. If I went above 110kHz on the primary and added the secondary it would tend to add about 12kHz. This resulted in poor output. I also noticed that when output was high moving my hand towards the top of the secondary would increase spark length to a point. Once i got too close one large spark (about 6") would jump out and the whole thing would switch to ~120kHz.

I figured out the problem, its a bit of a stupid one. I was using this set up for audio modulation of a plasma flame from a TV flyback. My goal being that I want to audio modulate the big Jacobs ladder, but starting small. I have 2 inductors in this set (no center tap primary) up so to audio modulate I added a third inductor between the ZVS inductors and the positive power supply. I added a coil (musical input coil) to this inductor that I pump an amplified music signal through and it works well. I got the idea in my head to short out this musical coil. When I did the out put suddenly doubled and become very noisy. Then it hit me, my inductors are too small *face palm* I build up bigger ones and removed the musical input inductor and suddenly the thing works like a charm. I'm using IRFP260N so I cant take the input voltage higher than about 60VDC, but I get about 12" sparks off the thing and it pulls about 13amps AC.

I have some new transistors I want to use to take to a higher voltage but so far the are worse than the 260's. I'm using D3S099N65B-U and IHW40N120R5XKSA1. The IGBT actually works better than the MOSFET so far and got an output close to the 260's. However, I'm noticing a massive voltage spike on the IGBT when dialing up the voltage and I need to check again but the MOSFET gate seems to be developing a spike in the negative direction. The 260's dont get such a massive voltage spike and the gate stags flat and only goes into the positive.

More testing to do!

[davekni]

I'm using IRFP260N so I cant take the input voltage higher than about 60VDC, but I get about 12" sparks off the thing and it pulls about 13amps AC.

60Vdc to a ZVS oscillator results in PI * 60V = 188V peak.  Startup transients will be even higher.  No room to increase voltage without changing parts.

I have some new transistors I want to use to take to a higher voltage but so far the are worse than the 260's. I'm using D3S099N65B-U and IHW40N120R5XKSA1. The IGBT actually works better than the MOSFET so far and got an output close to the 260's. However, I'm noticing a massive voltage spike on the IGBT when dialing up the voltage and I need to check again but the MOSFET gate seems to be developing a spike in the negative direction. The 260's dont get such a massive voltage spike and the gate stags flat and only goes into the positive.

Yes, IGBTs can work.  Delayed turn-off causes issues, however, forcing the opposite IGBT's reverse diode to conduct momentarily.  Then the diode's turn-off charge causes spikes.  Fast IGBTs with fast internal reverse diodes are best.  Here's my IGBT ZVS project:
https://highvoltageforum.net/index.php?topic=831.msg5491#msg5491

Higher voltage FETs have a different issue.  Forward voltage drop gets too high to keep the opposite FET off (keep Vgs below threshold voltage).  This problem gets worse as FETs heat up.  Forward drop increases and threshold voltage decreases.  Here's a circuit I've designed and simulated, but not built yet.  It is for high voltage IGBTs, but the same idea would work for high voltage FETs, to bias the gate voltage lower than opposite drain:
https://highvoltageforum.net/index.php?topic=1658.msg12681#msg12681

[Andrew321]

Hmm... I would order more parts to try out on of your circuits, but I just received a batch and I try to space it out to avoid over spending. Especially around this time of year lol.

I think for the IGBTs I'm going to add a top load to try and slow it down more. Its operating at 95kHz as is, but if I can slow it down more, that should help. How much, I guess we shall see!

Also, for the higher voltage MOSFETs if you don't mind checking my reasoning to make sure I'm understanding this correctly that would be much appreciated! So the forward voltage drop is dictated by the Rds and the current flowing through. Higher voltage MOSFETs tend to have a much higher Rds so for a given current the voltage drop is higher. The gates are discharged through fast diodes and the MOSFETs so the combined voltage drops should not be greater than the gate threshold of the MOSFETs. If it is this mean that the gate is not sitting at 0V but at some higher voltage.  Correct? This would explain why the D3S099N65's I got work at lower voltage but as I turn it up it failed. The current rises with the voltage and so does the voltage drop. Does this sound about right?

The fast diodes I'm using have a drop of 1.4v.
The D3S099N65 at 10amps would have a drop of about 1V
Combined is 2.4V the min threshold for the gate is 2.3V and typical is 3.2V
So depending on the part and how hot it gets, this would not work.
The IRFP260N at 10 amps would have a combined voltage drop of 1.8V and a min threshold of 2V So it works.

What doesn't make sense to me is that the 260's I've taken to 30 amps, maybe 40, so at best a combined drop of 2.6V. I have never had them fail and while their max threshold is 4V I just doubt that I just always get fets with a threshold always above 2.6-3V. Am I missing something here?

[davekni]

Also, for the higher voltage MOSFETs if you don't mind checking my reasoning to make sure I'm understanding this correctly that would be much appreciated! So the forward voltage drop is dictated by the Rds and the current flowing through. Higher voltage MOSFETs tend to have a much higher Rds so for a given current the voltage drop is higher. The gates are discharged through fast diodes and the MOSFETs so the combined voltage drops should not be greater than the gate threshold of the MOSFETs. If it is this mean that the gate is not sitting at 0V but at some higher voltage.  Correct? This would explain why the D3S099N65's I got work at lower voltage but as I turn it up it failed. The current rises with the voltage and so does the voltage drop. Does this sound about right?

Yes, nice explanation!

What doesn't make sense to me is that the 260's I've taken to 30 amps, maybe 40, so at best a combined drop of 2.6V. I have never had them fail and while their max threshold is 4V I just doubt that I just always get fets with a threshold always above 2.6-3V. Am I missing something here?

No, you are not missing anything.  It is not a hard sudden failure threshold.  FETs with Vgs slightly above threshold conduct a small current.  That generates some extra heat, but doesn't immediately fry anything.  That heat increases die temperature, which increases RdsOn and decreases threshold voltage.  At some point the excess heat becomes enough to cause thermal runaway.  Increasing temperature causes increased "off" state conduction, causing increased heating, etc.  That is why many of the Chinese induction heater circuits use high speed fans to squeeze a bit more power from a given IRFP260 ZVS circuit.  (Also, diode forward drop at the relatively low gate pull-up resistor current is well lower than spec voltage drop at full rated current.)

[Andrew321]

Thanks Dan for all the helpful advice! I will continue to test and share progress

Also thanks to Mads for his helpful advice at this website. Specifically the bit about the top load.

Here is the progress so far!