Mazilli ZVS flyback driver

WARNING!: Working with electricity is dangerous, all information found on my site is for educational purpose and I accept no responsibility for others actions using the information found on this site.

Read this document about safety!


The Mazilli ZVS flyback driver is well-known throughout the high voltage community for its simplicity and ability to deliver 20-50kV at high currents for a flyback transformer.

I build this circuit almost a year ago on a vero board, but it kept blowing the thin traces due to high currents flowing. I eventually put the project in a box and forgot all about it.

Inspired by the point to point soldered designs Myke from the forums often uses, I tried to make something in that manner, not as pretty as his work though.



The MOSFETs used need a voltage rating about 4 times higher than the supply voltage and a on-resistance below 150mΩ

5 + 5 primary windings are suitable for voltages between 10 to 40 VDC, at higher voltages additional windings will be needed. Experiment with the number of windings to improve performance. Too few windings will result in excessive heating and too many will result in reduced power output.

A MMC is made from 6 capacitors to avoid excessive heating in a single capacitor.

This driver will push as much power as it can, so be sure to use flybacks that can handle the abuse if you want it to live.



Voltage supply 35 VDC from a rewound microwave oven transformer.
MMC 0.66 uF from series string of 3x 2 275VAC MKP X2 capacitors in parallel .
Power consumption 400 Watt.
Longest arc 100 – 110 mm long white arcs



15th may 2009

I have now rebuild the driver using 2.5mm² / 14AWG wire for a good current ability, larger heat sinks and a MMC to avoid as much heating as possible.


16th may 2009

I found 4 different flybacks from my collection, among these are a 1980’s Bang & Ollufsen television flyback. A small flyback from a photocopier. A  flyback from a 1990’s portable television, it is without screen and focus resistor networks. A flyback with rectifier tube from a black & white 1950’s television.

The pictures with long arcs about the size of  100 – 110 mm was made with the 1980’s Bang & Ollufsen flyback transformer.


It was well worth it to rebuild this driver. It can now handle long run times with little heating despite pushing out around 400 Watt  of power!


104 Responses to Mazilli ZVS flyback driver

  1. Daniel Cuneo AKA wannabegeek on says:


    I am a physics student with some experience with electronics. I have ordered parts for the ZVS
    driver you are using.

    I have a question about the toroidal inductor. I have some ferrite toroids laying around I pulled from computer
    monitors while getting their flybacks. I don’t know how to look up their characteristics to calculate the inductance. In the past I tried using a bridge and Oscope, but that has not worked well for me.

    Is there a rule of thumb that will get me moderate enough accuracy to make a good circuit ?
    The flyback I have choosen puts out a strong arc with my 555 driver. More so than another larger looking FB.
    My MMC kinda sucks since I didn’t figure what to buy until after I spent some money, so I need to break (use) what I have first; 50, 2KV ceramic 0.01uF caps, 10 series = 1nF with a max of 5nF.

    As well, I don’t have 2W resistors but do have 1W resistor….hoping that will work enough to get started
    especially since I can only draw 5 A from my supply.

    I’d love to make my own power supply. I have tried to make a small one for practice but the loading is confusing me. I started with 35 Vpp and by the time it as rectified and had a large cap in parallel, it was down to 3 VDC. I would prefer to make it regulated. I had an electronics lab which built a Wilson mirror and a BJT voltage. Do you have a link to a good schematic or can I use yours ?

    thanks for reading all this…

  2. Mads Barnkob says:

    Hey wbg

    I would advice you to get one of those real cheap LCR meters off of Ebay, they have a good accuracy compared to their very low cost. A LCR meter in invaluable in electronics anyway, you need it!

    Salvage some MKP X2 capacitors from a pc power supply or anything else that uses them for input filtering, if you have no money to spend on these very common and cheap capacitors, your MMC is far from enough with only 1nF.

    Get a variac for a “cheap” variable supply, a variac will take any abuse up to 300% of its current rating for short runs of a few minutes, with a nice lab power supply you will risc damaging it powering up circuits like these that can have some heavily back EMF through the circuit.

    Kind regards

  3. Pingback: Video of My High Voltage ZVS Flyback Driver - Laser Pointer Forums - Discuss Lasers & Laser Pointers

  4. saeed haroun says:

    I am an engineer working in the maintenance of machines for laser
    I want to know if you have a high-volt flybacks transformer, such as in the image
    it is used in 80w laeer power supply
    How much is the price
    Please reply as soon as the maximum

  5. mostafa says:

    why fast diode use?

  6. mostafa says:

    why fast diode use in orginal mazili zvs driver ?

  7. mostafa says:

    how use Schottky diode Substitute fast diode ?

  8. Alex says:

    Hi, could I substitute the IRFP250N’s for IRF540? I do not think there is too much of a difference between them but just though I would double check.



  9. Mads Barnkob says:

    Hi Alex

    You can use the IRF540, but only at lower voltages, about half, than you could with the IRFP250N

    Kind regards

  10. Alex says:

    One more thing, is the large MMC vital? or can I use a smaller arrangement of odd value caps to make 0.68uF or perhaps even a single cap just for now?


  11. Mads Barnkob says:

    You can use anywhere between 0,1 to 1,5 and properly other values too, but depending on voltage ratings, capacitance etc they will generate a great deal of heat loss, which i why I made that arrangement.

    It will also change the resonant fequency of the circuit, so experiment with different values and see what performance you get :)

  12. Alex says:

    Thanks, Sorry to keep bugging you with questions like this but I like to make sure I am getting suitable components before I buy them.

    For the inductor is this a suitable one note that it IS a 2 pin inductor and the picture is just for illustration.


  13. Mads Barnkob says:

    Make your own, its cheaper and can withstand higher current.

    If you use a good ferrite core you can get away with as few turns as I, if its a yellow iron powder core from a atx psu f.ex. you need 70-100 turns I think.

  14. Alex says:

    Thanks, So the inductor on that page is not suitable for this type of purpose?

    The problem with making my own is that I do not have an LCR meter to measure the inductance of a homemade inductor.

    I think I can get hold of an old ATX power supply from a relative who I will ask if they will let me have it sometime during the week. I can also salvage the heatsink and cooling fan from it.


  15. Mads Barnkob says:


    You should buy one of these:

    Its a cheap hong kong LCR meter, but to the money and amateur use its accurate more than enough :) I have this exact meter and it have served me well, tested up against my fluke meter there is really no difference.

    What I meant by making your own is that you learn something about inductance, tedious winding work and getting parts for free, if we wanted we could buy it all, but where is the fun (or where are all my money!) in that? :)

  16. Alex says:

    I have now got the parts and have come up against something I can not find any answers for on the internet, the pin-out of the two IRFP250NPBF MOSFET’s I have purchased.

    Which pin in the gate, drain and source?



  17. Alex says:

    Thanks, I figured it out in the end and have got it up and running.


  18. Mads Barnkob says:

    Congratulations, that is a fine output you have there, not too current heavy, so might be useful for some high voltage applications that does not require a large current.
    Try to up the voltage and you will get some really fat bright arcs :)

  19. Alex says:

    I just tried it on 24v. Wow! It made a big difference, much hotter, thicker and longer arcs. My chicken stick electrode sparked like a sparkler if I let it get hot enough.

    I should have a video up in the next day or so as I am going to wait until it gets dark and film some arcs with the lights off.


  20. Alex says:

    Video of my ZVS driver running on 24v input is uploaded now.

  21. Mads Barnkob says:

    I am looking forward to see it on 36V :)

  22. ciprianwiner says:

    Hi Mads, You have a really nice setup there 😀 i will try to buy some better mosfets because mines just dies IRFZ44N and other that are non suitable for this application so my question is if i can use: IRFP240 or IRFP450 or IRFP9240PBF don’t know which(IRF450 is my choice because of the price) please let me know which of them is the best for the job.
    Regards Cip

  23. Ricardo says:


    Can be used IRFP460PBF?

    Diode MUR860 ZVS circuit?

    I await your return
    Thanks Ricardo

  24. Mads Barnkob says:

    Hello Ricardo

    I can understand from your parts in question that you want to run at higher voltages, it can be done but you might run into some practical problems like…

    1) If you run the circuit from a variable power supply, you will initially draw too much current at low voltage and your components might not be able to handle that.

    2) You need some voltage overhead for your MOSFETs to survive, using the IRFP250N MOSFET rated for 200 Volt, I would say that about 40 Volt is good safe limit. Thus for the IRFP460BF it would be around 100 Volt input, maybe even less.

    3) Higher voltage also means that you will be running a higher Volt/turn for your primary coil wound on the flyback transformers core, this will eventually fail if its too high, you should use 8+8 windings or more. More windings will also mean reduced power output, but your MOSFETs will survive instead.

    In the end there might not be a big difference in the output from going for high voltage as you have to take your precautions to insure its safety.

  25. Alex says:

    Hi Mads.

    YouTube have recently gave me monetisation features which enables me to make money from adverts on my videos.

    I have enabled them on a few of my videos already but though I would get your permission first before I enable them on the video I have embedded in my comment above. Since I followed your component list in the schematic to make this circuit I feel I should ask first before I enable adverts since you came up with what components to use.

    Just let me know in the comments here, I will not be offended in any way if you do not want me to enable adverts on the video I uploaded above.



  26. Mads Barnkob says:

    Hi Alex

    Its is infact Vladimiro Mazilli that deserves all the thanks for this design, just a heads up on ads, it takes years to get a payment when you are just a private person 😉

    Kind regards

  27. Alex says:

    Thanks for the reply. I will leave the revenue sharing for that video turned off then, or at-least until if the video ever became very popular (which I doubt it ever will).

    Oh I know that it was designed by Vladimiro Mazzilli and I have given him credit in the description. (Any idea what year he came up with it?).


  28. LAKSHMI says:

    can is use the 2N3055 instead of IRF250at same 20vDC please reply me i am about to do this because this circuits gives good performance rather than other circuit i am waiting for your reply…….

  29. Mads Barnkob says:

    No you can not, if you look at the datasheets for the 2N3055 transistor and the IRFP250 MOSFET you will see that there are considerably differences in their specifications and they are not even the same type of transistor.

  30. Alex says:

    I finally got round to trying mine out on 36v input,

    lol I originally had the same problem as you did on that photocopier flyback, my primary coil wire was too thin and started smoking!

  31. Mads Barnkob says:

    Hey Alex

    Good results you got there :)

    Keep experimenting!

    Kind regards

  32. Usman says:

    Does anyone know any replacements for MKV capacitors? I can’t find them anywhere…
    I am currently using newark to purchase my parts, “”.

  33. Usman says:

    I went through my stash of capacitors and found a 2500 volt, 1 micro-farad capacitor. Would that work just as well?

  34. Mads Barnkob says:

    Hey Usman

    1uF would work fine, but watch the temperature of the capacitor.

    Its a MKP not MKV capacitor, its short in Europe for metallized polypropylene capacitors

    Kind regards

  35. I was wondering what the difference is between the irfp9240pbf and the irfp9240pbf-nd ?

  36. Mads Barnkob says:

    Hey Robert

    nd often means No Diode, but you should check the data sheet to be sure about differences on parts. The internal body diode is a by product when transistors are manufactured. But they can also be without.

    Kind regards

  37. widarr says:

    Hello! I want to point out another important matter with this driver: Use a good (large) filter capacitor on the bridge rectifier. If you hear the 50/60 Hz humming, you have not enough capacitance. The transistors will have a greater power loss (and heating) when turning off while there is still some power left in it due to the mains halfwaves. I got one FET to spectacular blow up while experimenting with this ^^;

  38. Shelmak says:

    Umm, a question… can it be runned with a 555 audio modulated flyback driver?

  39. Mads Barnkob says:

    Shelmak, please describe your question is more details, I am unsure what you are asking me.

  40. hanghuutri says:

    Hi everybody, I not work in physics but interested in flyback circuit. I tried the circuit of Vladimiro Mazzilli. I don’t have a couple of IRFP250, so I replace with a couple of IRF640, but then they’re broken and the mosfet dead, then IRF840 Rectifier, they’re broken and dead still, fast diode I used is SUF30J, the voltage input is 12v and 5Amp. Can anyone help me…

  41. hanghuutri says:

    With the ZVS flyback circuit of Vladimiro Mazzilli, if I place the 47-100uH in wrong direction, does it make the circuit broken and the MOSFET dead???

  42. Mads Barnkob says:

    Hello hanghuutri

    12VDC is too little, if the MOSFETs are driven with a too low voltage you can not control them properly and they will short circuit. You need to raise your voltage to over 20VDC to get reliable operation if you use the same source for gates and power.

    The MOSFETs used need a voltage rating about 4 times higher than the supply voltage and a on-resistance below 150mΩ.

    You can not place the inductor in a wrong direction.

    Kind regards

  43. hanghuutri says:

    Thanks Mads Barnkob, I will try it.

  44. hanghuutri says:

    Hi Mads Barnkob, a question, I see the circuit use mosfet without rectifier, I’m going to use the IRFZ44V, it’s seem compatible with the input voltage I use, but it’s a rectifier mosfet, is it ok?

  45. hanghuutri says:

    I mean that I see the zener diode symbol in the IRFZ44V’s schematic diagram, is it compatible for the circuit?

  46. Mads Barnkob says:

    Hello hanghuutri

    The reason for a external diode is that the complementary diode in the MOSFET is only a by-product from manufacturing the MOSFET. The ratings of these body diodes can be inferior and thus a faster external diode is used.

    Kind regards

  47. hanghuutri says:

    Thanks Mads, I finished the circuit and it’s run ok.

    But when I use the circuit as the induction heater, it’s failed when I winding the primary coil in disc-shaped.

    When I put a metal disk which need to be heated too near the primary coil (in disc-shaped), the mosfet get hot quickly and burned out while the metal disk doesn’t even get warm.

    When I winding the primary coil in tube-shaped, the induction heating is ok. Does anyone know what wrong?

  48. Mads Barnkob says:

    Hey hanghuutri

    Have you tried to calculate the different resonant frequencies you get with different work coils? That could be your problem. If you have cylindric coil that works, try to match its inductance and frequency in designing the pancake coil.

    Kind regards

  49. hanghuutri says:

    Hi Mads, I’m sure that I used the pancake coil with the same inductance.

    By anyway putting the metal too near the primary coil may be the reason (I think so, not sure),

    because I tried with a little longer distance (about 2cm) between the metal disk and the primary

    pancake coil, it work ok, but the heating is too slow, and performance is not acceptable… cry

    whole night 😀

    I have some another questions, after searching on Google, I don’t find one.

    1. What the maximum frequency this circuit can obtain? (I don’t have an oscilloscope :p).

    2. I’m going to use this circuit to generate powerful microwave, is this possible?


  50. hanghuutri says:

    Hi everybody again, please help me with replacement problem:

    On the success circuit: I used two mosfest IRF640, the fast diodes I used is SUF30J ( The coil L1 I used toroid coil (take from PC power supply), with inductance is about 100uH. It’s work ok, the mosfets are even not warm.

    On the last circuit I tried, I replace IRF640 with IRFB4020PBF which has the same Vds as IRF640 (200 Volt) except that it has lower total gate charging, lower on-resistance – (

    And the fast diodes I used is schottky diode MBR10100G (Vrrm 100 Volt) instead of SUF30J (Vrrm 600 Volt) – (

    Anything else is not changed, and the last circuit failed, the mosfets were too hot. Too low total gate charging cause the circuit fail? or fast diode I replace is problem? Any idea?


  51. hanghuutri says:

    Oh God, sorry everybody, I find out that I picked up resistors with wrong value.. new mosfet (IRFB4020PBF) and new diode (MBR10100G) is work ok.. my mistake.. 😀

  52. Andrea says:

    Seen some questions here.

    I’ve built many of those little pesky drivers, and it’s really a versatile little thing. I’ll add some useful information.

    As Hanghuutri asked, maximum frequency… I drove this circuit into the MHz range, using two IRF520 (very low gate charge) with low gate resistors, stopping at about 15V supply voltage. It was resonating on an air core, something like 3 turns total, with a few nF worth of capacitance as Tank (Polypropilene, good quality KP caps). That circuit worked at about 1.6MHz, but the gate signals were a little curved, not a good square wave. Well, even with such fast mosfets, the RC constant between charging resitors and gate charge was perhaps less than a tenth of the total period.

    Now on further info on this circuit.

    1) Gate Resistors. Adjust them, depending on your frequency, supply voltage, and gate capacitance. Their role is to charge up the gate fast enough after the feedback diode stops conducting, and therefore ram the mosfet into a saturation state as fast as possible. Many schematics offer a value of 470 or 680 Ohms. That is good for normal operation, 12-40V and up to perhaps an hundred KHz. If you scope the gates, you’ll notice if you need a lower resistance, as they’re charging lower than expected. On the HF design, I was running 120Ohms resitors on 15V supply, made of 4x 470r 1w in parallel, per gate. That kept the rise time decently low.

    2) Feedback diodes. If you want to ram your frequency UP, they have to be fast. No need for them to handle a buckload of current, because there won’t be much. They’re just discharging a gate, as fast as possible, into the opposite drain which has gone deeply negative. Normal 1A ultrafast diodes are OK.

    3) Fets or IGBT’s. As said, keep their breakage voltage at least four times above your supply voltage. five to be sure. Meaning, if you are crazy enough to build this circuit to withstand the 230/240V mains, don’t bother with anything smaller than 1200V rated transistors. Keep their connections to the capacitors and the trasformer’s primary as short as possible, and make sure there’s room for all the current they’ll have to handle. If you want to go crazy, some additional protection might come handy. This thing is literally a grim reaper for power transistor, when misused or brought close to its limits. Mov’s, perhaps an antiparallel ultrafast diode, or even some neon glow bulbs to protect from overvotlages might come handy.

    4) Resonant capacitor. Don’t ever underestimate this thing. pick good quality capacitors, made to sustain high RMS currents. Which means, possibly polypropilene, or anything similar. Don’t go for the crappy MKT caps, they’ll overheat and die quickly. Also there, use caps rated with voltages much higer than your supply, building them into MMC’s if necessary. Keep connections short, use copper foil if needed to handle large currents between the tank cap and the primary winding.

    5) Transformer’s primary. This circuit, makes the primary of your transformer into an LC parallel resonant circuit, which means, a buckload of current flows in. Use an adeguately thick wire, and expect it to melt mercilessy when you ram up the power. If you want higher frequencies, litz wire might come really handy, due to the evil skin effect lurking around in any RF design. Also here, keep connections between the tank cap and the primary as short as possible, as it’s where all the reactive energy is pushed around.

    If you have any more quesitons, just reply here. sooner or later I’ll try to reply. I’m build a solid state tesla based on this circuit, so I’ll post somewhere when it’s done. perhaps on 4HV.


  53. Mads Barnkob says:

    Hey Andrea

    Thank you for the input, would you mind if I added this information to the article itself?

    Kind regards

  54. Dylan Burger says:

    Hi , I have a problem, I am using IRFP460 MOSFET’s , byv27e ultrafast diodes , 17v zener diodes , and a .68uF microwave cap, the driver runs my Flyback at 5 amps 20v and 8 amps 50v but when I connect to my sparg gap tesla coil the current jumps to 20A before tripping my fuse , it regularly burns out zener’s , is it because 17v os not high enough, or because the tank voltage rises too high ? (almost 100V) . Does anyone have suggestions?

  55. Mads Barnkob says:

    Hello Dylan

    Could you post a link to a schematic, it sounds like you have a short circuit in order for the current to rise that high.

    Kind regards

  56. Dylan Burger says:

    Thanks for the reply Mads, here is a link to the scematic from uzzors2k: ,
    I have modified this cuircuit a bit ( higher rated diodes and zeners for higher voltage and IRFP460’s , I dont know if the zener diodes keep blowing up because of the input voltage or the tank voltage…
    If there is a way to prevent that ( ive heard of people putting a 7812 regulator in for the gate drive ) I’d like to do it because it gets e pensive when the diodes pop and my mosfets short out…

  57. Mads Barnkob says:

    I think you should try a separate supply for the driver voltage.

  58. Dylan Burger says:

    Ok, thanks Mads, I will try that!

  59. Andrea says:

    Are you sure that your flyback transformer isn’t giving out DC? SGTC’s are normally working from an AC supply, unless you run them in dc-resonant charging mode ( Also, in a normal SGTC, the transformer is shorted when the spark gap fires… meaning, you need a transformer (like a NST) capable of running short-circuit. This would be like when you draw arcs to ground with the flyback, rocketing power consumption into the high skies.

    Zeners should not blow up, perhaps are you overvolting your input? a fixed supply can be good, better 15-18V than 12 (for much quicker gate charging). Also, keep connections between the mosfets drains, resonant capacitor, feedback diodes and primary winding as short as possible, as you don’t want any possible stray inductance around that might shift your neutral potential up blasting your gates apart.

    Besides, there’s been some experiments by our fellow Tesla coilers on the italian mailing list, building a direct 4kW ZvS sstc, works quite well, almost to full line power (~200V ac, 280V dc in). Standard circuitry, 1200 or 1500/1600V Igbt’s and some crazy large mmc to hold the primary resonant currents. But, before jumping on such project, it’s rather advisable to come around the circuit and learn how to avoid it from tearing apart.

  60. Dylan Burger says:

    Thanks for the info on DC res. Charging Andrea, I will try that, my tesla coil is indeed a DC coil, my imput voltage on the driver is far higher than 12 V ( close to 50VDC) and I am probably overvolting the input, but I will fix that with one of the 12V windings on my XFRMR so that the gate wil have about 15v DC and the tank will still have 40-50VDC supply, I think I should use 600v ultrafast diodes to keep them from shorting out when the tank voltage rises, my driver has no problem with an 8Amp draw , so im only worried about the zeners dying and killing the ‘FETs.
    Thank you for your info on the DC Tesla coil charging.

  61. Andrea says:

    Well, 50V are a heck of a lot of juice pushing into the little thing. If you kept resistors the default value, you are indeed frying your zeners rather quickly, mostly due to overheat. 470/680R are designed for 12-15V, at 50V or so you got so much current that your diode will just go boom and crack open. Then, the gate gets 50V, and well, bad fate for it.

    We did redesign this driver with a Flip-Flop / gate driver idea, if I find the schematic I’ll post it.

    I’d say it’s good to go for a separate power supply for the gates, keep in mind the RC constant (R as the charging resistor, C for gate capacitance), and that your gate will be up to 63% of supply voltage after Tau = RC, and to 86.4% after 2Tau. Keep that in mind to obtain a decent rise time on your mosfets, and switch them on quickly enough. But, as you said, I bet you’re around 20-50kHz, so plenty of room for quick gate switching.

  62. Dylan Burger says:

    Its a lot, but it’ll handle it if you dont push it too much. I look forward to building the flip-flop :)
    Thanks for the info on gate charging, and I will keep that in mind because I want to build a small sstc using the driver , if I cant find a simpler way, I was thinking of using a d209l transistor wired for feedback oscillation (I think thats what its called) and run it from some high-ish voltage source to get oscillation. I dont think this is the place to discuss that further though, (it is a mazilli driver page after all)
    Your guess was pretty good, I run it at 60Khz for max power and I run it up to 160Khz for lower power operation, Thanks for the help and I look forward to a better ( and cheaper ) future for my driver !

  63. Dan Smith says:

    Great posts, very good feedback from a lot of experimenters. I was about to build a vandalism deterrent for my car, for the 2nd, full alarm I’d have a capacitive discharge type circuit that would give maybe 10 jolts for a second, then pause for a second or two, then repeat… and a ignition coil with high voltage output connected to car frame and a chain/link that drops down and contacts earth. Something to jolt the evil doer. This with it’s long arc ability interests me a lot. Any ideas on making a 12 volt, lower power (more primary turns) that will give a good spark but far less, non-lethal, current unit? How about using an ignition coil for the flyback transformer? It should give the high voltage but at less current. Ignition coils are not center tapped primary though, so a different version driver would need to be used, more like a Meisner oscillator, with a resonant capacitor across the primary. Any ideas… Comments… Feedback? Thanks.

  64. Mads Barnkob says:

    Hey Dan

    Do not use high voltage as a weapon or defense system, unless you buy a certified system. You will only risk putting the DIY community in a bad light if you hurt someone and tighter regulations might hit us.

    What good is a car that electrocutes people, which pisses them off and they throw a brick through the wind shield?

    Kind regards

  65. John says:

    Hi Mads, Someone who is “electrocuted” is more than pissed off, that person is dead! Electrocution is always fatal. An electric shock isnt always fatal but would certainly make me unhappy! John.

  66. Mads Barnkob says:

    Hi John

    Hehe, I can see that results in a conundrum, I hope I can get away with excusing that English is my 2nd language :)

    Kind regards

  67. ntoskrnl says:

    Hey Mads

    Continuing the discussion here:

    I applied five-minute epoxy to the flyback several times to avoid arcing, and eventually ended up cutting off all the unnecessary pins and covering the whole bottom in epoxy. However, the flyback only produced a tiny, barely visible spark anymore. So I figured it was friend internally, and obtained another transformer.

    This one produced a really nice arc that could be stretched to 15cm and it was hot enough to melt a steel electrode (1500°C) in ten seconds! The bottom pins did arc one time so I cut off the pins and covered the bottom in five-minute epoxy again.

    After that we got around 10 minutes of runtime. The voltage and power of the arc were decreasing all the time, and now it was just barely better than the 2N3055 circuit. We noticed that the tank capacitors were bulging noticeably.

    They were warm, but not hot to the touch; perhaps 50°C. The tank is made of three 0.22µF 275V MKP X2 caps in parallel. I measured the capacitance of the bulging caps with my cheap Chinese LCR meter and it said 0.2µF for the bank of three caps, so they are obviously fried.

    I didn’t know MKP caps could bulge. I guess they just weren’t quite up to the job? What kind of caps would you recommend?

  68. Andrea says:

    Those oscillator are hellish. What did you power the circuit off? I suppose 30V the least, giving the results. Caps are DC rated 275V, not AC. The resonant system will place about three times Vcc on the tank capacitor, which means, 30V in, you get 100v rms on the cap. A golden rule on those things, is oversize. critically oversize.

    Same circuit as yours, at 40V, I busted a 3.3µF 400V capacitor with quite a bit of RMS current capacity. Try to solve your circuit with a simulator, estimating the series resistance at a low value… you’ll get tens of amps running through your poor capacitors, which most likely aren’t rated to survive long periods of such torture.

    Go for higher voltage caps, build a small MMC, or switch directly to rail polypropilene’s intended for H-Bridge or inverter usage. That will ensure (most likely) that your circuit won’t massacre any more your capacitors. And, most importantly, while doing this, keep connecitons short, and use copper plate if you need big currents (remember skin effect depth at your resonant frequency).


  69. ntoskrnl says:

    I’m actually powering this from a 48V 50A power supply. :)

    The caps I used are rated for 275V AC:

    But I guess the voltage tolerance is derated for higher frequencies. Also no mention of RMS current tolerance on that page.

    Do you have any specific capacitor models which you’ve had luck with?

    What is a typical resonant frequency for this circuit by the way?

  70. Mads Barnkob says:

    Hey ntoskrnl

    I agree with Andrea on this, the capacitors are not rated for 150VAC at properly some 100+/- kHz and the RMS current could be so high that they melt and bulge out.

    Take a look at the page about the Royer induction heater where this same circuit is described more in detail about ratings and some measurements.

    15cm sparks are also impressive and you are violently abusing a flyback transformer if you can pull some serious current from those 48VDC :)

    Kind regards

  71. ntoskrnl says:

    Okay, I was considering something from the Cornell Dubilier 940C series. Their voltage ratings are plenty enough, and according to the datasheet they are capable of around 10A RMS at 100kHz. They are also used in Tesla coils.

    The question is, how many do I need in parallel? They are pretty expensive so I don’t want to go overboard. I can get

    – 0.1µF 1.2kV caps for 4.5€ each
    – 0.22µF 2kV caps for 8.3€ each
    – But then, 1µF 1kV caps for 7.6€ each.

    I wonder if a single 1µF cap would do it? It’s rated for 500VAC and 12.5A RMS at 100kHz at 70°C.

  72. Photonned86 says:

    Nice article! I have a question for Andrea if you’re still around the forums… I’ve went about your methods on making the zvs operate at much high frequencies for air cores and everything seems to be working fine… 260n’s, 5nf, 4+4 turns primary, and 32 awg on a 2 foot secondary, im not sure what its frequency is but its getting into the higher areas… Well anyways my main question is that its working great and staying cold, I can draw arcs from the secondary to objects grounded to me for around 4 seconds and then it starts to die out. The voltage input is 35volts and my smoothing bank is around 80000uf with full wave. Also when I crank up the voltage to 50 volts the breakout gets bigger but becomes “out of tune?” I tried replacing the smoothing bank with 50vdc caps but the 35vdc bank seems to work out better. Are the fets dieing from to much current? Is this a known behavior that you’re aware of? Any suggestions from anyone is also welcome! Thanks in advance…

  73. Andrea says:

    What resistors you use to drive the fet’s gates? A cause of failure can be the zeners being fried due to overheating, and thus getting 50v on the gates for a short time, but still destroying the oxide layer inside your mosfets.

    A zener diode typically goes short circuit when overdriven in the typical reverse bias configuration, but, before going short circuit, its reverse voltage rockets up, until it behaves for a short time as an open circuit. After a few seconds, the diode cools down, and it goes into a short circuit stage. That time that passes here, is rather enough to break your fet’s junctions.

    If you want to rocket up this driver’s supply, i’d suggest a separated rail (perhaps regulated with a 7815 or 7824) to keep a constant voltage on the resistor – zener voltage drivers on the gates.

    There has been a remake of this driver with logic (flip flops) and gate drivers to avoid cross saturation and gain a better stability on the italian high voltage mailing list, I’ll take a look through it if I can find it, I’ll post it there.

  74. Photonned86 says:

    Sorry Andrea I forgot to mention I have 4 1k in parallel for each gate and an external linear adjustable source supplying 15 volts capable of 20 amps for them. I went thru the period of frying my zeners to the point of knowing every resistance of every part on this driver for it to be working xD. Should I be able to just lower the gate resistance values? Are the gates modulated after this is squared away with say a 555? Thank you again!

  75. Andrea says:

    The driver a friend on the italian mailing list put together uses a flip flop to ensure correct oscillation, and most importantly, a pair of UCC37321/322 heavy duty gate drivers to pull the gates up in record time and avoid switching losses. I’m heading to work soon, will check now if I can fetch the schematic from somewhere.

  76. Andrea says:

    Well, i’ve found the driver. All thanks go to Diego Barone, a fellow high voltage enthusiast whom redraw the original royer schematic adding some logic and stability circuitry.

  77. Mads Barnkob says:

    Hey Andrea

    Did you also see that he built a working unit from that schematic? I have tried building on myself with flip-flops and gate drivers, it worked but was very unstable and in the end it killed my MOSFETs. I scrapped it and would instead build a whole other driver topology for higher power.

    Kind regards

  78. Photonned86 says:

    Thank you so much Andrea. Mads I thank you for this forum too! I ended up going with wards mini sstc. It can be a monster if done right! But its just not enough!!! What is it, just cant get enough? Well i’ve decided to go full on dual resonant sstc. Thought maybe pll but why not skip and go drsstc! Never the less, the ZVS driver can accomplish loads of current in lots of different ways. Its worth seeing what all it can do. Not just center around a ferrite core. BUT we all know with a core this is the most stable way “so far” to have stable current at high voltage and very high frequency. Thanks again and to anyone else that ponders the world of hv “safely” and decides to share their discoveries!

  79. ntoskrnl says:

    I got the 1µF 1kV cap I mentioned earlier and assembled the circuit again. It worked great for a little while until it just died. One of the FETs was shorted.

    I replaced the FET, and added a 180V TVS in parallel with the cap to absorb any overvoltage ringing of the choke with the cap when the arc breaks (I found this situation possible when I simulated the circuit).

    Now I got a bit more runtime with nice 10-15cm arcs, but the same thing happened again – a shorted FET. The flyback secondary reads around 200kΩ in both directions; I’m not sure whether that was there before or if there are shorted turns now. The flyback did arc a little bit to the core on the inside.

    Are my FETs dying due to the inherent instability of this driver, or do you think they died because the secondary arced to the primary through the core?

  80. Mads Barnkob says:

    Hi ntoskrnl

    I looked back at your previous posts and could not see which MOSFETs you use, I would say that your supply voltage is low enough to not damage a IRFP250N as it would only see around 150VDC from a 48VDC supply.

    As you supply can deliver all the current the circuit properly is able to use, the MOSFETs might simply die from die overheating due to too high RMS current? Could you measure the inverter voltage, inverter current and tank current like I did on the induction heater? ( )

    You are getting some good results with 10-15 cm arcs and when it arcs to the core you are punishing that poor little flyback way beyond its V/turn ratings :)

    Please upload a picture of your setup in the comment, you can do that now if its less than 2MB.

    Kind regards

  81. Dylan says:

    Great job! Mine is running on 50 VDC and a 0.68 uF tabk cap, my switches are 600v 30 A igbt’s, ive blown up many flybacks since! Works great as induction heater when you get the coil just right. Constant current draw of 10A on the DC side :) mads, you have helped make a monster!

  82. Dylan says:

    Heres an image of the insides, it has 2 ups transformes wired in phase and series for +-40 AC, when rectifeid and filtered, its +-50vDC. Big ‘ ish heatsink, way overkill unless you use as an induction heater. 18v 1W Zenner diodes ( not sure of exact voltage) , and some kickasss Ultrafast diodes rated for 5A. Not sure of make or exect details either because i bought them on sale at a local electronics shop. Most important is all my cuircuit protection and a analogue Ammeter , up to 30A DC. Im thinking of using as a spark gap coil power source. I did not intend to make a very neat build as i was rushed by my exitement of blowing up flybacks :)


  83. Mads Barnkob says:

    Hey Dylan

    That sounds very interesting, very few people have had luck building this with MOSFETs or IGBT with a rating over 200VDC, due to differences in gate construction on higher voltage rated switches.

    Please tell which IGBTs you used and if you changed anything from the original circuit, except using iron core chokes.

    Kind regards

  84. Dylan Burger says:

    Ive had a quick look and, they are g20n60b3d IGBT ‘s, I have not modifeid the circuit in any way, its still a standard mazilli driver, ps: those arent chokes – they are the 2 x ups power supply transformers wired in series to double voltage output to +-40VAC :)

  85. ntoskrnl says:

    I am indeed using IRFP250N FETs.

    Neither the heatsinks or the transistors themselves heat up any noticeable amount, so I don’t think they are overheating. The FETs were also fine with another flyback, which died due to arcing.

    Unfortunately I don’t have an oscilloscope so I can’t take any measurements.

    Here’s a photo of my setup. The heatsink was an Intel stock heatsink which we cut in half. They are isolated from each other. The 200µH choke is pretty small physically, so I guess the core saturates at high DC currents; I’m not sure what effects that could cause.


  86. ntoskrnl says:

    Here’s a closer look:


  87. Dylan says:

    ntoskrnl, if i may suggest something – try a smaller cap to get the freq up a bit, in my experience, if you put a cap that size on a mazilli at 40+ VDC, its going to pop something, also , take a look at your zenner diodes, mine were the conventional 12v zenners untill i was fed up with continuously having to replace them ( it would literally burn the zenner in half) im using 16~18 v zenners, and, a bad zenner would cause a FET to fail
    Hope this helps, ive had lots of trail and error with this.

  88. Mads Barnkob says:

    Dylan have a good point in driving the MOSFETs harder with a higher gate voltage, but be sure to check the maximum zener clamping voltage so it does not exceed the maximum allowed gate voltage of the MOSFET. The IGBT Dylan refers to do also have a +-30V gate voltage pulse rating, compared against the IRFP250N only have a +-20V gate voltage rating. It can however be overdriven a little, to some 24V maybe.

    A lower switching frequency is better for the MOSFETs, they will spend less time in the linear region if the gate drive is a bit on the weak side.

  89. Himanshu says:

    Hey can I use 1n4148 for switching diode

  90. Mads Barnkob says:

    Hey Himanshu

    MUR1560 is a 15A 600V ultra-fast diode whereas the 1N4148 is a 0.2A 100V small signal ultra-fast diode. It is no good in this circuit with its low voltage and current specification.

    You should compare the continuous reverse voltage/current/recovery time when you try to find a substitute for the MUR1560.

    Kind regards

  91. Jeff Freeley says:

    Hello Mads,

    Thanks for all of your awesome work. Do you know if this circuit might work with a regular car ignition coil? I am looking for a way to create ozone generators and this would be a perfect option. To vary the output would you just change the voltage in?

    Thanks, I am relative newbie on this subject. : )

  92. Mads Barnkob says:

    Hi Jeff

    A car ignition coil us made with a iron core and might just saturate at the high frequency this circuit runs at with the usual component values. You would have to experiment with larger L and C values to bring the frequency down.

    You could also look at using a 2N2055 or TL494 driver instead.

    Maybe using flyback transformers instead could be a idea? They are easy to scavenge from old TV sets.

    Welcome to the high voltage hobby :)

    Kind regards

  93. Jeff says:

    Thanks Mads, I guess I am looking in the wrong direction for my needs. I am wanting to create a variable ozone generator circuit, so looks like ZVS is not a good way to go about it. There are a few 555 ignition coil circuits I can utilize, but also noticed some of the manufactures are using what looks to be some type of modified low output ZVS circuit because they are relatively silent.. (vs loud pulse of the 555 ignition coil drivers).. They are seem to be more efficient power wise..The potted one I have here looks to have 2 mosfets so that’s just my guess.. I don’t know enough about electronics to be able to figure it out, so my best bet is to stick with the old school 555 ignition coil driver circuits that do fit this application, or perhaps search for paid help from someone who specializes in this area of design.. : )

  94. Dylan says:

    Hi Jeff , From my experience the 555 flyback driver is the best way to go bothfor power and frequency adjustment ( for low power / high corona discharge).

    A zvs driver is very finiky and if you may not like to replace your mosfets every time you change something. Also it may be way too powerful for simple ozone generation.

    I have attempted a ion propulsion unit with both a zvs driver and a 555 flyback , and the zvs driver kept on arcing to my other terminal where the 555 kept a nice and steady corona discharge.

    An iron core tranfomer (ignition coil) runs at a low frequency(10Hz ~ 1kHz) -which is why it is so noisy- where the ozone generator you described seens like it would be ferrite core ( 20Khz +) which is why it is “quiet”.

    I would be happy to assist you any further if needed. But please note – I am no expert :)

  95. Jeff says:

    Thanks Jeff! That is helpful, I appreciate it. I will keep looking around to see what I can find, Cheers!

  96. Jeff says:

    Sorry, typo, I meant to say thanks Mads! : )

  97. ntoskrnl says:

    I finally got back to this. I rebuilt the circuit on a piece of stripboard, and attached the MOSFETs with screw terminals so that they can be easily replaced. No MOSFETS have died so far though.

    I tried with a lower voltage this time, first 24V and then 36V, rather than the full 48V. My SMPS based power supply doesn’t sag under load, so this is probably more like the unregulated transformer/rectifier power supplies many others use. I also avoided breaking the arc as the FETs usually died when I did that.

    The circuit works very well now and can melt the tips of steel electrodes. We had several five-minute runs as well, including one where we put the electrodes in a glass jar and taped it shut. The arc produces nitrogen dioxide, and after arcing for five minutes or so and letting it sit for a while, the air inside turn distinctively orange-brown.

    After long runs, the flyback secondary gets pretty warm, I’d say about 60–70°C on the surface of the transformer. The 2.5mm^2 primary also gets warm. The tank capacitor and 0.75mm^2 input choke get slightly warm to the touch, and the MOSFETs get perhaps a few degrees above ambient.


  98. Dylan Burger says:

    So , ive got myself interested in the ZVS driver again , and dig out my old build. I never really got decent induction heating power from it and was curious about what i could do to improve.

    It runs off of ~46 VDC and will trip a built in fuse if current rises above 10A on the DC supply. Im using IGBT’s ( which ive been told is a rare feat with the mazzilli driver ) rated at 600V 50A ( if memory serves ) .
    I have 18v Zenner diodes and i cant remember which the ultrafast diodes are.
    Tank cap is a .68 uF 1200v microwave capacitor.
    Work coil is (currently) 8 + 8 turns of 1mm magnet wire from a microwave tranformers primary.

    What could i do to improve performance and also get this thing to not destroy 50 bucks worth of IGBT when i bypass the tripswitch?
    (It can run a flyback indefinitely @ 8A with no heating to IGBT’s,and it can make a 8mm stainless rod turn orange within 18 seconds @ 9A ) Im afraid to push it harder without modification because of the instability i had experienced when building my first ZVS driver.

    I also want to remove the centre tap and add another inductor instead, for the ease of making a decent coil.

    Also , what can be done to increase output voltage of the flyback without pumping more juice into the DC side ? Ive seen folks with a driver 1/3 rd the size of mine running 150Kv arcs ( mine can currently put out 70 ~ 100Kv before it trips by tripswitch ).


  99. Mads Barnkob says:

    Hey Dylan

    If it is possible, try to use a primary coil with a larger gauge, or parallel some more magnet wire to have a lower resistance and lessen skin effect.

    Fewer primary turns will give you a bigger output, but also increase primary current and vice versa for more primary turns.

    More turns on the secondary side will also increase your voltage output.

    You could increase the capacitor size up to about 2-4 uF.

    The ZVS driver is by nature not very stable, it will suddenly explode from hardswitching large currents or overvoltage. Remember that the switches in a ZVS driver sees pi(3,14) x supply voltage.

    Kind regards

  100. Dylan says:

    Thanks mads.

    I think the main issue is my capacitor that is too small. Its a .68uF microwave cap, that gets fairly hot after a bit of running.

    I took the weekend and decided to rebuild it and totally bypassed the 10A Tripswitch.

    I wound a work coil with 7 turns 8mm copper tubing and am using high current wire on all connections , it will draw 25A now without a hitch @ 48vDC but im not able to heat anything up to the point where it starts to glow.
    The tank freq is 115kHz and tank voltage is around 100v.

    I am under the impression my input voltage is too high, but im also unsure of what freq the tank needs to be for more heating power.

    Im not sure how to measure tank current but i will do research on how.

    In flyback mode with a 5 turn primary it will draw 15A @48v and 10A @ 20v , produces thicker, hotter arcs now , but no noticeable voltage output
    ( i measure via air breakdown voltage, 10cm is the point at wich it starts to arc).

    Is lower tank freq better for induction heating ? ( i was thinking ~60kHz ).

    ^^This is a photo of my current setup

  101. Dylan says:

    My bad .. that was halfway through the build .

    Heres the full setup :

  102. Mads Barnkob says:

    Hey Dylan

    I can see that you are still using a microwave oven capacitor. You should replace that with some metallized polypropylene capacitors .

    Without water cooling of the work coil, be sure to watch that it does not get so hot that the plastic container melts :)

    Kind regards

  103. Dylan says:

    I modified the work coil , it worked extremely well, and then my microwave cap exploded XD.

    unfortunately it blew up my IGBT’s as well, ill be making an mmc and hopefully i can achieve decent output.

    the 50A high current cables hooked up to the cap got incinerated too,.

    Ill be using new 900v 51A igbt’s made specifically for SMPS power supplies, they should work well.

    If i destroy anything else ill post an update 😋

Leave a Reply

Your email address will not be published. Required fields are marked *