DRSSTC using a ZVS Driver not working well - need help

[BenTesla]

Hello I'm attempting to build a DRSSTC using a ZVS driver as shown here

However I get very poor spark output with tiny arcs about 1mm in length and can barely light up a florescent bulb. I have built my own interrupted ZVS driver shown in photos.



The circuit draws 13v at 180mA. I've been running the circuit for the entire time and components remain cool to touch. I have experimented with tuning tesla coil and believe primary is in tune with secondary. The circuit is interupted with a 10% on time at ~ 50Hz

Here's a photo of my setup. the timebase on oscilloscope is 1us/div


I'm at a slight loss as to why performance is very poor as.

• I think the system is in tune
• No component heating
• While power input is low compared to other peoples this should still be more than enough to get at least 1cm arcs. I built an SSTC driver for the same coil and that can do 3cm arcs at 30v 200mA input

I'm wondering if my ZVS driver is poor and that is why its not working well. When I ran it as an induction heater at low power there wasn't really any heating or increased power consumption but it does oscillate fine. If so can someone explain from the schematic why performance is poor

I'm using 2N7000 mofsets in parallel for my switching transistors I know that is not ideal but I used to use 2 IRF540N mosfets and performance was the same but I accidentally blew those up in an experement. I need to order some more power mosfets. The transistors are cool to the touch.

With the tank capacitor:
I'm using a 33nf tank capacitor with 5 turns on primary. For tesla coils with ZVS drivers people seem to use much higher capacitance maybe 100nf. Why is this. I think because as Xc = 1/wc increasing capacitance lowers reactance so more tank current. With larger capacitors starting the ZVS oscillation is very hard. I may need to use a higher current power supply.

Also my ZVS driver is running at about 300KHz

If anyone can help me get this working better I would be very grateful as I'm reaching the limits of my knowledge. Thanks

[davekni]

Your 13V supply must be high-impedance to get any oscillation.  ZVS oscillators of this type generally cannot be interrupted except with another switch (FET or IGBT) interrupting power input.  At the start of each interrupter enable pulse, both FETs turn on.  They will not start oscillating unless they manage to pull their supply down to the level that a gate is turned off through a drain-gate diode.

Also, arc length is not linear with power.  Significant voltage is required to start any arc.

I recommend you learn LTSpice or other analog simulator.  It is easier to experiment with circuit changes in simulation than with real hardware.  In simulation (if you chose) and with real hardware, do have fun experimenting and learning!

[BenTesla]

Thanks for your reply. I think I need to learn circuit simulation and build this in LT spice as being able to easily see voltage and current in the circuit would probably help a lot as clearly some of my component values are not optimal.

I'm not sure why the supply must be high impedance as don't the 100uH inductors provide the necessary impedance matching for resonant voltage rise and for oscillation so the FETs can pull the supply low enough to turn of the gate. If i start with too low of a current, oscillation does not happen as both FETs turn on at the same time.

Also I know my method of interrupting is not ideal as when both transistors are turned off by the interrupter it causes voltage spikes because of the high di/dt through the inductors. as you said I need to use another transistor to interrupt power.

Finally In the video they said that for CW coils you shouldn't use a topload, why is that? I understand it makes breakout harder but it adds frequency stability and lowers the resonant frequency on the secondary which can be easier on the transistors.

[davekni]

I'm not sure why the supply must be high impedance as don't the 100uH inductors provide the necessary impedance matching for resonant voltage rise and for oscillation so the FETs can pull the supply low enough to turn of the gate. If i start with too low of a current, oscillation does not happen as both FETs turn on at the same time.

Yes, if the gate voltage is turned on fast enough, then the 100uH inductors will provide necessary impedance.  I see a capacitor to ground on the gate supply, so presumed it was turning on more slowly.  Couldn't quite make out the capacitor value.

Also I know my method of interrupting is not ideal as when both transistors are turned off by the interrupter it causes voltage spikes because of the high di/dt through the inductors. as you said I need to use another transistor to interrupt power.

You have a good understanding of the issues.  Rapid gate turn-off will cause a voltage spike, driving FETs into avalanche breakdown.  Slow gate turn-off will minimize the spike, but still dissipate the inductor's stored energy in the FETs.  If too slow, FETs will dissipate much more excess power as oscillation drops out while FETs are still conducting DC current.

Finally In the video they said that for CW coils you shouldn't use a topload, why is that? I understand it makes breakout harder but it adds frequency stability and lowers the resonant frequency on the secondary which can be easier on the transistors.

Since it's not my video, I can only speculate.  You can add a breakout point to the top or side of a top-load, so mitigate that one issue.  However, even with a breakout point, lower frequency requires more voltage to start an arc.  Since most ZVS circuits operate at lower voltage than H-bridges, they probably want a coil that works with lower voltage.

Also notice that their SSTC has relatively few primary turns located in close proximity to secondary.  That helps generate more secondary voltage given the low primary voltage.  If you learn LTSpice, you can make some very simple coupled-inductor circuits (TC primary and secondary) with capacitors to simulate top-load capacitance and ZVS resonant cap.  Use JavaTC to calculate inductances and coupling factor for different physical designs.  Put those into LTSpice to see how they perform (upper and lower pole resonant frequencies etc.).  BTW, ZVS oscillators tend to lock to the upper pole.

If you want to venture farther into simulation, you could also learn FEMM.  I did just over a year ago.  Never too old to learn   That way you can model more unusual variations such as winding the primary with wide copper foil instead of round wire.  Helps increase coupling factor for low turn-count primaries (ie. 1-3 turns), which would be ideal for ZVS-driven coils.

[BenTesla]

However, even with a breakout point, lower frequency requires more voltage to start an arc.  Since most ZVS circuits operate at lower voltage than H-bridges, they probably want a coil that works with lower voltage.

Very interesting so because a top load lowers the frequency they have removed it for better impedance matching. That had never occurred to me but poor impedance matching between my driver and coil is probably contributing to poor performance.

Why is it that lower frequency requires more voltage to start an arc. Is it because induced e.m.f is proportional to frequency by Faraday's law of induction?

Also notice that their SSTC has relatively few primary turns located in close proximity to secondary.  That helps generate more secondary voltage given the low primary voltage.

I understand this is also to improve impedance matching as with our ZVS driver we have high current and low voltage but we want low current and high voltage so we make use of transformer step up action.

relatively few primary turns located in close proximity to secondary

So because my primary is far from the secondary it makes achieving that high frequency harder on the ZVS driver because the primary has a greater inductance (f = 1/2*pi*sqrt(LC)). so increasing the coupling coefficient would help mitigate this issue as the primary would have a smaller diameter.

How else can changing the coupling coefficient affect the performance of my Solid state tesla coil especially with the ZVS driver? I've read Richie's piece on it here: https://www.richieburnett.co.uk/operatn2.html#coupling but that is for a SGTC so I'm not sure how relevant it is for my situation.

Finally I don't understand what you mean by the the

upper and lower pole resonant frequencies

. Is it this: https://www.richieburnett.co.uk/operatn2.html#splitting
How / should I take it into account when designing my Tesla coil for the ZVS driver?

With regards to simulation I'm going to start learning LTSpice to help me optimise component values and maybe the open source KiCAD. FEMM also looks very interesting for designing the actual primary coil.

Thanks very much for answering my questions as I've only been learning power electronics for about 3 months (before, I did audio electronics). Its extremely helpful being able to get expert advice on these systems as only so much stuff is documented on the web. Hopefully when I go to university I will be able to meet more people who know this stuff.

[davekni]

Why is it that lower frequency requires more voltage to start an arc.

I'm not an expert in air electrical breakdown.  My simple view:  On say a positive half-cycle positive air ions are created.  On the next negative half-cycle, those nearby positive ions increase local field strength.  At lower frequency, the ions travel farther from breakout point before the next half-cycle, so have less effect on local field strength.  Same is true for negative ions on negative half-cycle followed by a positive half-cycle.

How else can changing the coupling coefficient affect the performance of my Solid state tesla coil especially with the ZVS driver?

There is a lot of discussion about coupling coefficient on this forum.  I'd suggest searching here and other TC sites.  For SSTCs such as this, higher coupling is almost always better, as long as primary coil doesn't get so close to upper parts of secondary that corona or arcs form between the two.  SGTCs are different.  Too much coupling reduces performance of SGTCs.

Is it this: https://www.richieburnett.co.uk/operatn2.html#splitting
How / should I take it into account when designing my Tesla coil for the ZVS driver?

Yes, good link for pole splitting.  Probably not of particular importance for your SSTC.  Just expect the operating frequency to be a bit higher than the resonant frequencies of the hopefully-matched primary and secondary frequencies.

Thanks very much for answering my questions as I've only been learning power electronics for about 3 months (before, I did audio electronics). Its extremely helpful being able to get expert advice on these systems as only so much stuff is documented on the web. Hopefully when I go to university I will be able to meet more people who know this stuff.

You are welcome.  Have fun learning!  You are off to a great start, and clearly know quite a bit already!

BTW, if you are planning to major in engineering, I'll suggest that you also take as much physics as possible.  Physics provides a better theoretical underpinning for engineering principles.  That is valuable when confronted with unconventional design/debug tasks.

[Mads Barnkob]

However, even with a breakout point, lower frequency requires more voltage to start an arc.  Since most ZVS circuits operate at lower voltage than H-bridges, they probably want a coil that works with lower voltage.

Very interesting so because a top load lowers the frequency they have removed it for better impedance matching. That had never occurred to me but poor impedance matching between my driver and coil is probably contributing to poor performance.

Why is it that lower frequency requires more voltage to start an arc. Is it because induced e.m.f is proportional to frequency by Faraday's law of induction?

I think you are mixing up two different things here.

I guess they remove the topload, for CW mode, to get a lighter load on the driver. A huge topload capacitance is just more work for the system to deliver energy into.

So while a topload capacitance also lowers the frequency, its not because of the lower frequency to run a certain voltage, its to maintain CW operation on a lighter load with the same input voltage as interrupted mode. CW mode on a heavy load has made some of my SSTCs explode or burn

[davekni]

I guess they remove the topload, for CW mode, to get a lighter load on the driver. A huge topload capacitance is just more work for the system to deliver energy into.

For a given secondary winding (given inductance), top-load will lower frequency and therefore lower secondary impedance.  Is that what you mean by the top-load being "more work for the system to deliver energy too"?  That doesn't directly imply any more input power required, just a change in impedance matching to compensate.

Or, are you thinking of higher secondary current with top-load (for given voltage) and therefore higher I^2R losses in secondary?  Secondary I^2R losses are on the order of 100W for my 3kW uninterrupted SSTC (50cm diameter top load, 100kHz).  That would be a large fraction of total power for a low-power coil.

[Mads Barnkob]

I guess they remove the topload, for CW mode, to get a lighter load on the driver. A huge topload capacitance is just more work for the system to deliver energy into.

For a given secondary winding (given inductance), top-load will lower frequency and therefore lower secondary impedance.  Is that what you mean by the top-load being "more work for the system to deliver energy too"?  That doesn't directly imply any more input power required, just a change in impedance matching to compensate.

Or, are you thinking of higher secondary current with top-load (for given voltage) and therefore higher I^2R losses in secondary?  Secondary I^2R losses are on the order of 100W for my 3kW uninterrupted SSTC (50cm diameter top load, 100kHz).  That would be a large fraction of total power for a low-power coil.

I am not sure I can explain it right

The system with more capacitance on the topload, takes longer to charge before breakout is achieved. With or without topload in this case, I see it as with or without breakout point. The system has to work harder before long streamer and steady state is reached.

[BenTesla]

I guess they remove the topload, for CW mode, to get a lighter load on the driver. A huge topload capacitance is just more work for the system to deliver energy into.

That makes a lot of intuitive sense

That way you can model more unusual variations such as winding the primary with wide copper foil instead of round wire.  Helps increase coupling factor for low turn-count primaries (ie. 1-3 turns), which would be ideal for ZVS-driven coils

This also looks very interesting. I will defiantly look into FEMM after learning LTspice.

How I'm going to make this work well
Following advice here I'm going to build a LTspice simulation and get good resonant current through the ZVS tank and then improve the impedance matching between primary and secondary by optimising the number of primary turns and the coupling coefficient as we've discussed here. The next step is for me to do this and then I'll post the results here if I have any problems. Does this sound like a good plan?

One more thing: What about using a transformer as shown below to improve impedance matching between primary and secondary. Would this work well, it seems like quite a good idea to me?


I got the idea from this induction heater tutorial: http://inductionheatertutorial.com/ where they had the opposite problem to me of trying to convert high voltage from the half-bridge into low current.

Thanks for answering all my questions as I now feel very informed about the design choices for this type of coil. I feel confident on why its not working and how I can improve it.

[davekni]

The system with more capacitance on the topload, takes longer to charge before breakout is achieved.

That is clear.  Would think it to be important only with short interrupter pulses (as typical for DRSSTC) where time to breakout is a significant fraction of pulse width.

With or without topload in this case, I see it as with or without breakout point.

Without breakout point makes a huge difference.  Depending on top-load geometry/smoothness, I think voltage before breakout can almost double without breakout point, for 4x energy in top load capacitance and 4x power dissipation in resistive losses.  The 4x energy could be handled by a long enable pulse and appropriate impedance matching.  The 4x losses has no easy work-around, requiring 4x power from drive electronics.

Not sure if these example ratios are reasonable, but I guess that top load could be significant for power required.  If top-load cuts frequency in half (4x capacitance), restoring frequency requires 1/4 secondary inductance.  Half the turns at twice the wire diameter gives 1/8th resistance.  For given breakout voltage (presuming break point is added to top-load), primary current is 4x (4x capacitance), so secondary I^2R loss is 2x = (4x * 4x * 1/8th).

Does this sound like a good plan?

Yes, sounds good.

One more thing: What about using a transformer as shown below to improve impedance matching between primary and secondary. Would this work well, it seems like quite a good idea to me?

Your modeling may show that optimum TC primary turn count is 1 or 2.  It is hard to get good (high) coupling factor with 1 or 2 turns without using wide copper strips.  In that case, it might be simpler to wind a more typical 6-10 turn primary and use a transformer for matching.  Transformers are never perfect, so will add some loss and parasitic series inductance.  However, a transformer may be more practical than a 1-turn primary.