NGTB40N120FLWG-Based Tesla Coil

I need a SPICE model for the NGTB40N120FLWG insulated gate bipolar transistor. I can't find one online, and I don't know how I'd go about finding a similar IGBT that has a similar function. I'm using LTSpice to simulate a solid state Tesla coil, and I need to know how much current would be passing through five or six of these IGBTs with the emitters in series with collectors and a 12V pulse at the gates. The IGBTs would be discharging a 2nF capacitor charged to around 5600 volts for a small test coil. The capacitor would be much larger--possibly as large as 100nF--for the coil I'll be using for my science fair project. If the current would be too much for these IGBTs, I could place three in series and another serial triple parallel to those while also reducing the voltage to around 2800V. I could also try reducing the gate voltage with a voltage divider. If anyone could provide me a SPICE model for an IGBT similar to the NGTB40N120FLWG, that'd be great. It'd also be very helpful if someone could predict the amount of current that will pass or give me tips on this part of the circuit.

IGBTs can be stacked but transformer isolated gate drive is used. I've done this with MOSFETs and it is not a project for a beginner.

I recommend a Tesla coil design using a single IGBT. Even without an IGBT SPICE model a good simulation can be had. The circuit can be explored quite nicely using a voltage controlled switch or a MOSFET in place of the IGBT. The IGBT will turn on fast - 50 ns - but turn off over a period of a few hundred ns with a 10% "tail current." Driving an IGBT fast takes much less gate drive than a MOSFET.

Three things that limit the usefulness of an IGBT: tail current, high forward voltage drop, and high thermal impedance (due to the small die compared to an equivalent MOSFET).

I've seen several recommendations for IGBTs over MOSFETs for Tesla coil use, and I've already ordered the IGBTs.. Also, On Semiconductor, the producer of the NGTB40N120FLWG, has an application note that goes into paralleling of IGBTs, and they only recommended a gate resistor in series with both gates. If gate voltage oscillations occur, they suggest adding a resistor between each gate and the input signal and possibly adding a resistor between the signal and those two resistors. I haven't seen anything about putting IGBTs in series. Is that what you were referring to?

About the speed, it doesn't matter much to me. The input signal will only go up to 20000Hz for audio use, and frequencies up there are of seldom use anyways. I'm hoping to go far beyond that for my science fair project, but it's not necessary. The on time could be on the order of microseconds, so switching time shouldn't be a problem. The tail current may even help me drain excess resonating current that would make audio less....audible.

The forward voltage drop really shouldn't matter when I'm putting upwards of 1000V through each device.

Duty cycle will be relatively low to reduce heating and promote some oscillation on the primary side of the circuit. I've also read that using IGBTs instead of MOSFETs in Tesla coils reduces power dissipation because of less switching loss at low duty cycles.

In the end, the circuit doesn't really have to be all that precise. My only big concern is putting too much current through the IGBTs. I can use the datasheet to predict current of a single IGBT, but I'm not sure how to go about applying those figures to IGBTs in series.

Sorry, I thought when you said "emitter in series with collectors", meaning the IGBTs are stacked in series.

Your SPICE simulation can be done adequately using the LTSpice voltage controlled switch. The method is shown below.

If you'd like to use an active device in LTSpice, but cannot find an IGBT model to import, the STW11NM80 FET will work. Seven of these FETs have the ON resistance of one of your IGBTs and require about as much gate charge.

The IBGT gate must be switched in

Wow, I have never read anything about making sure the IGBT switches fast enough by allowing it to pull enough current to charge the gate quickly. That's really shocking and is gonna change a lot in my designs. If I were to control the gates from an op amp or 555 timer, would it be sufficient to just have a class AB BJT amplifier at unity gain to buffer the voltage from the op amp or 555 timer? (provided my voltage source can source the necessary current in the necessary amount of time)

Jeeze, I was going for a simple but extraordinary science fair project, but this is getting more and more complicated as I go on. For simplicity, do you think I should use these IGBTs or the MOSFETs that I was having trouble simulating in the other post you helped me with? I found a second one and two more similar MOSFETs, so an H-bridge can be made if an H-bridge would be more appealing than switching the primary circuit to ground or using an IGBT or MOSFET in place of a spark gap.

I've been dumbing down my design because of all these small details that would cause huge problems. Now, it's just a 555 timer with a selectable frequency at a constant duty cycle. I'll attach it to this post. All I need is to add an additional output at the same frequency and duty cycle but offset by half the frequency to control an H-bridge. I'm clueless as to how to go about making the second output and could really use some help. I typically only use BJTs, MOSFETs, 555 timers, and op amps. If you could help me come up with something using those, I'd really appreciate it. I'd be willing to get an inexpensive IC for it if it comes down to that, but I'd much rather work with what I have and am used to.

Here are some application notes on IGBT gate drivers.

http://www.onsemi.com/pub_link/Collateral/AND9052-D.PDF

http://www.semikron.com/skcompub/en/AN-7004_IGBT_Driver_Calculation_rev00.pdf

IGBTs are forgiving of high current and I think your choice is better than MOSFETs.

How would I go about finding the derivative of the collector voltage with respect to time at turn off? Would that be determined by the collector's voltage source and parasitic impedance? Also, what would you say my switching frequency would be limited to? I'm hoping to be able to choose from at least five preset frequencies without going below 100Hz.

With sufficient gate drive the IGBT will go from full ON to approximately 10% ON (almost off) in 140-400 ns at 600 volts and 40 amps, depending on temperature. See page 2 of the datasheet.

The dv/dt at turn OFF (25 deg C die) averages 600V/140ns = 4300 V/us.

http://www.onsemi.com/pub_link/Collateral/NGTB40N120FLW-D.PDF

For the condition of 600 volts and 40 amps the total switching loss is specified as 5.8 mJ. If you switch 1000 times per second the switching loss is 5.8 Joules which is 5.8 watts. To the switching loss we add the conduction loss. This is the forward voltage X current X duty cycle. For example, 3.5 volts X 40 A X 10% = 14 watts. The total loss is 5.8 + 14 = 20 watts.

I hope you bought extra IGBTs. It seems to a law of the universe that bringing up newly designed power electronic circuits consumes silicon. This is one way we learn. It's more fun when your employer is paying for the silicon.

If only my high school and/or college would fund my studies and projects. I'd progress in my knowledge so much more quickly. I've inquired my calculus teacher, algebra II teacher, and electronics/physics teacher about help with this stuff. All three are electrical engineers. The calculus teacher did circuitry for torpedoes some several years ago. None of them could help me. The electronics/physics teacher lent me his collegiate electronics textbook from 20 years ago, and that was no help. The only information I didn't already know that was contained in that book was about digital electronics, such as CMOS circuits and digital processing. You've been more help to me in a single post than each of them combined. Anyways,

I'm running into issues simulating the circuit with the MOSFET you mentioned before as a replacement for the IGBTs. If I used the voltage-controlled switch, how would I go about accounting for parasitic capacitance? The capacitance values listed in the datasheet are labeled input capacitance, output capacitance, and total capacitance. How could I convert that into gate-emitter and gate-collector capacitance?

I'm also trying to run a couple of IGBTs in series to increase voltage handling so that I could use a pair of microwave oven transformers and a rectifier as a 1400VDC power supply, but if I recall, you suggested a gate isolation transformer. I don't have any high frequency 1:1 transformers for that, and I really doubt I could make some that would be anywhere near ideal. In spite of that, I'm still trying to simulate it to see if it would be possible. I've added some kind of voltage sharing with a few resistors and capacitors, but I don't feel like it would prove to function as cleanly (cleanly is also used loosely here) when put together in reality. I'll attach the schematic at the end of the post. The 100k resistor would actually be 10 10k 2W resistors in series. The 10 ohm resistors are also rated for 2 watts. The 23.5k resistor would really be two relatively large 47k resistors with unknown power ratings in parallel. The 4.7 ohm resistors will have to be acquired and might be 1/2 watt resistors to save money, as I don't think their average power dissipation will be too great. Speaking of power dissipation, would 20 watts really be a problem for these IGBTs? They're rated for much higher wattage and would be attached to heat sinks. I've only ordered 6 of them. My original intention was to use 2 for this project and then use 4 for an H-bridge later on. I have a nasty habit of ordering parts before finalizing the circuit. Anyways, I read that someone burnt through roughly 400 IGBTs before finally getting a working series connection between them. That's really putting some doubt in this plan, so I might move on to simulating an H-bridge and/or using a voltage doubler or tripler from the line for a 340VDC or 510VDC supply. The main problem I'd have with that would be creating the second pulse with the same frequency and roughly the same duty cycle that would be roughly halfway between pulses from the first pulse source.

Power dissipation and efficiency aren't a huge concern of mine, as long as the circuit is still reliable and would output measurable spark lengths. Spark length is going to be an essential dependent variable in the experiment. The nonessential (not even quantifiable without expensive equipment) dependent variables would be spark color and spark shape.

Using the voltage controlled switch the IGBT output capacitance is modeled as a capacitor across the switch. The capacitance is is the IGBT Reverse Transfer C of 180 pF plus the Output C of 240 pF for a total of 420 pF.

The switch is not quite the same as an IGBT but it gives you a way to experiment with your circuit.

I've seen schematics of stacked MOSFETs with resistive networks to couple the gates but they were low speed circuits. I can see how to stack IGBTs so the bottom one is driven and it in turn switches ON the top IGBT, but I don't see a simple way to supply a fast (high current) gate drive to switch the top IGBT OFF.

The gate drive transformer method uses small transformers that only need to support a pulse of 1 us. The transformer secondary is coupled to the IGBT gate through two back-to-back 10 volt zener diodes. When the transformer primary is quickly pulsed to +20V the secondary pulses +20 V and charges the IGBT gate through the 10 volt zeners to +10V. The transformer drive is returned to zero slowly so the secondary does not swing much below 0 volts. The zeners keep the gate from being pulled to 0V. To switch the IGBT OFF the transformer is driven to -20V thereby pulling the IGBT gate to -10V.

I've never stacked FETs or IGBTs but know about it in detail because another engineer and I built and sold 15 kV pulsers using stacked MOSFETs. He designed the power electronics and I did the control circuitry.

There is a learning curve with stacking IGBTs and I would avoid it if possible. Reducing the microwave oven transformer voltage is more of a sure thing. That can be done on the 120 VAC side using a Variac or with a series pass circuit. The Variac is nice and simple and allows adjustement from 0 to full voltage. A series-pass circuit can be made using a MOSFET, two resistors, and a bridge rectifier.

I'm not having any success attaching a schematic. Can you walk me through the procedure of getting a schematic from LTSpice to here?

I just took a screenshot, pasted it into Microsoft Paint, then trimmed it down.

I looked around and most solid state Tesla coil drivers are push-pull. I found a couple of flybacks.

It looks like you're building a flyback. With a blocking diode in series with the IGBT the IGBT switches ON and charges the Tesla primary inductor to some current. The IGBT turns OFF and the energy stored in the inductor is transferred to the Tesla primary capacitor. The IGBT voltage will fly up above the DC supply voltage. The switch voltage will then swing down by an (almost) equal amount. It can swing below ground and that is why a blocking diode is needed as the IGBT will not support a reverse voltage.

The 20 ohm resistor limits the IGBT current as the 25 nF capacitor charges. This circuit is simple and is safe for the IGBT. 75% of the energy is dissipated in the 20 ohm resistor. Changing it to 10 ohms half the energy is dissipated in the resistor, assuming the secondary load does look like 1.5 k ohms.

Here's the voltage at the bottom of the Tesla primary.

Doesn't flyback refer to a flyback transformer? No matter. From the circuit you showed, that primary waveform is perfect. I may want it to dissipate a little more quickly so that peaks upwards of 20kHz would be more differentiable from a non-dissipating waveform. I'll try to simulate that with my anticipated tank circuit values as soon as I get home.

I did a little algebra and trigonometry and decided to try to emulate the IGBT as well as I could. I think it might be my best choice for simulation unless someone decided to make an accurate SPICE model of it in the past few days. It's fairly inaccurate at a lower Vge. What do you think of it?

Vpulse should be Vtest. I'm not sure why I added R1. Just in case it matters, it passes ~14A instantaneously when Vtest returns to 0V.

Would the current being passed by an IGBT be the same at different collector-emitter voltages? If so, my model is way off. Maybe a voltage-dependent current source would be more applicable.

With this being a loosely coupled circuit the IGBT current is dependent on the primary LC. With an inductor the current is found by:

V = L(di/dt), di = V/L dt

The current will ramp up from zero and should be terminated before the IGBT limit is reached.

But your circuit has a parallel capacitor and that presents a problem for the IGBT. Without a limiting resistor the capacitor charging current is limited by either the IGBT turn-on time or by the IGBT saturating. The energy dissipated in a resistive circuit charging a capacitor is equal to the energy stored in the capacitor. That is W = 1/2 V^2 C. It can be better to dissipate this in a series resistor than in the IGBT.

I don't think IGBTs have the avalanche energy capability that a MOSFET does. And the IGBT you choose does not specify avalanche energy. So, you do not want the flyback voltage to fly up to where the IGBT breaks down. Notice that my circuit has a supply voltage of only 200 volts yet it flies back to 1200 volts. To protect the IGBT the standard flyback clamp can be used. It's shown in the attachment.

C2 charges to some positive voltage above the +200V rail. The value of C2 is chosen so that if all of the energy from the inductor is dumped into it it will charge up by perhaps 10% of the DC voltage across it.

For 20 amps thru 100 uH the energy is 20 mJ. Dumping this into 1 uF charges it 200 volts. So, the clamp capacitor should be over 1 uF. R2 is sized to keep C2 charged to some voltage, perhaps 800 volts.

D1-3 must be very fast recovery 600 volt diodes. Tr of

So should I also include the diodes in series with the IGBT shown in the other schematic you posted? Or should I use only the parallel diodes without series diodes, such as in the schematic you just posted? With the schematic you posted before but with my expected values, the output voltage across the topload was adequate (25kV peaks for .2 coupling coefficient, 170V source, and 10kHz 5% duty cycle switching) but the power across the resistor was too high for what I have available. Also, LTSpice shows the diodes passing both a positive and a negative current. That shouldn't be happening with a diode, should it? I'm using 1N4007 diodes in the simulation because that's what I have handy and would like to use. The .model statement is:

*SRC=1N4007;DI_1N4007;Diodes;Si; 1.00kV 1.00A 3.00us Diodes, Inc. diode
.MODEL DI_1N4007 D ( IS=76.9p RS=42.0m BV=1.00k IBV=5.00u
+ CJO=26.5p M=0.333 N=1.45 TT=4.32u )

The IGBT model I made with a voltage-controlled current source, capacitors, and a diode (the only parameter I put in the diode's .model statement was the typical forward voltage given by the datasheet) also shows negative power (which is expected because of the internal diode), but LTSpice only calculates average power and not effective power. I'm not worried about the power going through the IGBTs, though, because the 9.6 ohm resistor network I made (24 ohm 5 watt resistors) is taking a great deal of the power for the IGBT.

Speaking of those resistors, each one is dissipating 5.33 watts on average, according to LTSpice. They're rated for 5 watts each. Would adding cooling fans and separating them a few inches from each other be enough to go beyond the power rating?

The 1N4007 has a very slow reverse recovery time and will conduct for some time in the reverse direction. Diodes labeled "ultra fast" would be good in this circuit.

The clamp circuit (diodes, resistor and cap) protect the IGBT from going into avalanche. The series blocking diodes (ultra fast) are to allow the circuit to swing negative.

The advantage of the push-pull IGBT circuit is that it is self clamping. But, it requires a higher voltage power supply and a floating gate drive for the top IGBT. The flyback circuit needs a clamp but operates with a lower supply voltage and needs no floating gate drive. Like any other circuit there are tradeoffs to make with the Tesla Coil. I would be more comfortable starting with a flyback. Note that I've built flyback converters and push-pull forward converters but I've never built a Tesla Coil.

I have played music through a carbon arc though using a 15 kW constant-current plasma power supply (the Advanced Energy MDX-II). It was set to 25 amps and the audio signal was injected into the error amp. The volume had to be set right so that it wouldn't drive the current to zero and extinguish the arc.

I'm thinking that to play music on a Tesla Coil the primary current needs to be modulated. To do that a current sense resistor (0.05 ohms for example) can be placed in series with the IGBT source to provide feedback to an error amp. Another way - one that does not require closing a feedback loop and avoids the trouble of stabilizing the loop - is to feed the gate from an amplitude-to-pulse width circuit. More amplitude gives more inductor charging time and more current.

If I were to use parallel strings of diodes to increase current capability, should I add a small resistor to each string and then have one power resistor in series with the IGBT? Or should I omit the one resistor and use meidum-sized resistors in series with each string? The effective resistance would remain within the same range, of course.

I'm confused as to which value capacitor and resistor I should use for the clamp. The voltage across the capacitor varies with the switching frequency (obviously), but it also jumps down 10 to 100 volts from where it was at at what looks like 18ms intervals. I'm using 3 1N4007 diodes in series with a 2.2uF capacitor and a 47 kohm resistor in parallel with the capacitor. I'm thinking it could have something to do with the diodes being slow.

What diodes would you recommend for this setup? The RMS current across the clamping 1N4007's is over 1A. The diodes in series with the IGBT (10 2-diode strings in parallel and each in series with 24 ohm resistors) aren't going over .5A RMS, but they go negative, so I'll have to get new ones for that, too. I may change that to 5 strings, each in series with a 2 24 ohm resistor string because 2.4 effective ohms seems a bit too little. 9.6 effective ohms should suffice, right? I'd check the power dissipated by the IGBT, but LTSpice only gives average power, not RMS, and the power swings negative as often as it goes positive.

I'm not too concerned with making the Tesla coil play music at the moment. I'm just going for specific frequencies, each with the same duty cycle. When I'm done with science fair, I'll swap that timing circuit out with a rising-edge detector circuit (some kind of simple CMOS schematic I found). I'm going to make audio files where one channel is a square wave with 50% duty cycle and the other channel is background. The square wave would go into the rising-edge detector, which would shorten the on-time to a certain length set by a capacitor and resistor, and that pulse signal would feed into the IGBT. The other channel would go into a speaker system that is at a safe distance from the Tesla coil. I'd be open to try out different kinds of signal modulation to feed into the IGBT gate, of course.

The MUR460 is a cheap, "ultra-fast" diode that's in the LTSpice catalog. Would that be sufficient? I tried simulating my circuit with those, and the topload voltage peaks at exactly 27kV with 10kHz 5% duty cycle switching. That's pretty low for the 22.5" x 6.625" secondary I have in mind and the 31" x 4" topload I have made. Other problems I'm having are keeping the power dissipation of the resistors below 5 watts and getting the clamping capacitor to the 800V you suggested. In the schematic I'm attaching to this post, the capacitor charges up to 493 volts. The only power resistors I have at hand are a 510 ohm 50 watt aluminum-encased resistor and 10 24 ohm 5 watt resistors. LTSpice calculated the average power dissipated by each resistor as 6.45 watts. I could lower the input voltage, but then that would accentuate my problems with low voltage across the clamping capacitor and at the topload. I'm open for input voltages from 0 to 500 as I could use one of the IGBTs I ordered to switch a simple variable SMPS. Outside of that, I could do certain multiples of 2800V with microwave oven transformers. I have four, so i could do 700V, 940V, 1400V, 2800V, 5600V, 8400V, 11300V, and multiples of those with capacitor-diode voltage multipliers.

This schematic gives a simulated ~44.6kV peak at the topload, ~812V across the clamping capacitor, 4.66W dissipated by the 24 ohm 5 watt resistors, 1.94W dissipated by the 10 ohm 2 watt resistors, and less than 530V peaks blocked by any of the diodes. I'll rustle the air around quite a bit with some fans to hopefully prevent damage to the components. The 170V source is delivering 134.38W on average. That means the remaining power dissipation should be around 50W from the IGBT because it's the only other considerably resistive component, right? I'll use a fairly large heatsink and forced-air cooling directly on the IGBT and its heatsink. Could I replace D1-D15 with 1N4007 diodes? The RMS current through the MUR460s is simulated to be 882.51mA.

As long as the MUR460 die does not exceed 125 deg there's no need to parallel them. Without a heatsink it can dissipate nearly 2 watts in free-air.

I would not use 1N4007 because not only do they have significant reverse recovery time but the "forward recovery" is slow. I measured one once and under the test conditions it was 1 us. Forward recovery is when a forward voltage is applied and there is a time lag before conduction begins.

You might want to bring the circuit up with reduced HV and monitor the IGBT collector voltage. Then adjust the clamp resistor as needed.

Have you got an oscilloscope probe that's up to the job?

On something that's not going to ship to a customer I don't mind pushing parts hard. A homemade Tesla Coil is in that category. So, I'd not object to running semiconductor die at 150 deg C, running resistors at their power ratings, and running capacitors with no voltage derating.

The reason I paralleled them was because they'd be exceeding 4A on average if not paralleled. I recently changed the schematic so that only two strings of three are in parallel. I'm hoping to not have to use more than 10.

I'm a high school junior who has taught myself everything I know about electronics starting late spring this year, so no, I don't have an oscilloscope. I was considering asking my high school's electronics/engineering/physics teacher whether or not he had an oscilloscope I could use to accurately tune the primary side. If he has one, I'll also ask to use it to adjust the clamp resistor.

I'll be putting up to 100 half-watt 15 kohm resistors in series and parallel to achieve the necessary power rating and resistance, so adjusting the resistance should be as easy as moving an alligator clip or two to different spots. How would I go about adjusting it? Start low then move up until the IGBT collector voltage peaks at a certain value? And would I do that with the whole thing put together or with certain parts of the circuit removed?

On a side note, can the clamp capacitor have a capacitance that's too high? I don't have high voltage capacitors with capacitances in the lower uF range. I'd hope to be able to use 2 100uF 400V and 2 470uF 200V capacitors in series. The other option I could go with is paralleling a few microwave oven capacitors.

I would bring up the voltage then while monitoring the collector voltage. The voltage can be monitored at the DC side (diode cathode) of the clamp using a DVM.

After running for a minute or two I'd shut it down and feel components with a finger. This is my quick-and-dirty temperature probe. If you can touch it for only a few seconds it's about 70 deg C. If it's over 100 deg C wet a finger and by the sizzle you can tell if it's close to 100 deg C or if it's REALLY HOT.

The thermal resistance of the MUR460 is about 70 deg C/watt. So, it must be limited to less than 2 watts or about 2 amps unless it has a heatsink. Contrary to popular belief diodes of the same kind can be paralleled without resistors. One won't go into thermal runaway and hog all the current.

You're doing great and I'm surprised you're new to electronics. I can see you going into power electronics as a career.

I'd still be using a series-parallel resistor network to get the right resistance and power dissipation. I might as well set it up with the diodes within it just to ensure thermal runaway doesn't happen. The collector current of the IGBT with an effective series resistance of 5.4 ohms is 4.33A, so each diode with two strings in parallel would ideally pass an average of 2.165A. That should be fine with a large amount of ventilation, right?

Yeah, I plan to major in electrical engineering I'm just not sure which area I'd focus on. I like working with power electronics and high voltage, but I also dabble in audio-related designs such as amplifiers, signal processing, and audio spectrum analyzers.

Cody, I think it's great that you'll be concentrating on analog electronics. With the high edge rates (fast dv/dt) even digital has become analog and this specialization is called Signal Integrity.

I've done just about every type of analog for years. I don't do microprocessors or code.