Build an Audio Modulated
MOSFET Tesla Coil!
aka: Dr Hankenstein
Component Test Video Here
Completed Project Video Here
This project operates at lethal voltage and current levels! Recommended for advanced, experienced builders well versed in electrical, high voltage and RF safety practices. Author assumes no responsibility or liability. Experiment at your own risk.
Not too long ago when one wanted to build a Tesla Coil, there weren't too many options. Originally you could follow Nikola Tesla's lead and build a spark gap coil. In the 1930's experimenters began building Tesla Coils with Vacuum Tubes. They were of the regenerative type that are still very popular today. These tube coils are often referred to as "VTTC's" (Vacuum Tube Tesla Coil). It was reported that there was even a large VTTC that was audio modulated as early as the late 1930's at the World's Fair! Not much changed over the next sixty years. Pulsed Cathode (Stacatto mode) was pioneered in the early 1990's and has gained much popularity. I audio modulated a dual 833C coil a couple years back using fellow ham radio Steve Cloutier's cathode modulation topology. I used a single CM600 IGBT as the modulator. The circuit worked well, except the amount of real estate required for power supplies, biasing, neutralizing, etc. was incredibly daunting. Fortunately the last ten years have seen incredible advances in Solid State Tesla Coils. Many top notch experimenters have brought a variety of different designs to the table. Nice thing about most solid state coils is that you don't need a fork lift to move them around. Bad thing about solid state is that poor design, construction practices, etc. can result in a big, expensive bang. Many new and often inexperienced builders will purchase tons of parts, even buy kits that may not have very descriptive schematics, etc. only to be discouraged and give up. This is my first MOSFET design Tesla Coil and I was fortunately blessed to have Chris Hooper (Dr.Spark) as my mentor. Hopefully this article will help clarify some of the myths that surround the world of solid state coils. I will describe various suggestions and tips that helped make my coil a success.
The schematics herein are a collaboration of designs from Chris Hooper, Steve Conner and Uzzor. I recommend printing the schematics for reference. It is OK to use the schematics / photos for hobby / educational purposes. Commercial use is strictly forbidden. The project is layed out in various steps for ease of understanding.
(Click to Enlarge)
Shown below are the basic components of the Tesla Coil
Left Photo: contains the necessary electronics for driving the Tesla coils' primary circuit. In the forefront is the PLL oscillator and MOSFET driver assembly mounted to a heatsink. Above that is the 1/33:33 feedback CT (CT-1). Across the top from left to right are the low ESR electrolytic capacitor, MMC capacitor bank (C in the tank circuit), power mosfet driver board (mounted on heatsink) and HV bridge rectifier (mounted on smaller heatsink). These components combined make up what used to be known as the "spark gap" in earlier non solid state coils.
Right Photo: primary and secondary. Primary is a flat pancake design, secondary is a cone shaped coil with 950 turns of #31 AWG enameled wire on a 9" tall by 7"at the top and 8"on the bottom form. Note the 8" x 2" corona ring on top with 3" brass stinger. Resonance of the secondary assembly runs at about 138khz at full load.
By convention, the Tesla community typically distinguishes the above components in three categories as follows:
1. All the electronics, electrolytic capacitors, power rectifier, driver board, feedback CT's, power output bridge, etc. is often referred to as the "ENGINE". The engine provides the proper power and frequency to the Tesla coil's primary circuitry.
2. The primary coil and capacitor is referred to as the "TANK". It should be noted at this time that this particular Tesla coil has what appears to be a conventional "tank circuit"...but it's not! This will be discussed in greater detail later.
3. The secondary is occasionally referred to as the "RESONATOR". The resonator will often have some sort of toroid or sphere on top. This is also known as the "Top Load". The top load mainly serves two purposes. It acts as a "gradient shading ring" which helps keeps high voltage corona away from the winding thus reducing the effects of corona "eating away" at the winding, particularly at the high voltage end of the coil. In addition to protection from corona, the top load adds additional capacitance to the coil. The additional capacitance actually reduces the self resonant frequency of the resonator. The resonator and top load are electrically equivalent to a series LC circuit; the "L" being the resonator and the "C" being the top load. With this in mind you can see that the size of the top load has a direct effect on the resonant operating frequency of the coil. That is: The larger the top load, the lower the frequency.
FULL H-BRIDGE OUTPUT MODULE
Basically, when I started building this coil, I wanted it to be a full bridge (or H-bridge) design. This meant I would need four power mosfet's. I examined Steve Conner's, Steve Ward's and Uzzors designs in detail. Prior to this, I had never built a solid state mosfet coil before, so I decided to build the output with the biggest mosfets I could find. I purchased some APT50M60JN transistors off of ebay. These devices are rated at 500 volts at 71 amps each and have a large mounting tab that could be bolted directly to a heat sink. The layout required low impedance buss, transient suppressors across the Source to Drain conductors and a large snubber capacitor across the DC buss. The transient suppressors (TVS's) consists of two 1.5KE220E 220volt devices in series across each transistor. This would give 440 volts of transient protection. The snubber capacitor (C1) is a 10uf / 400 VDC, low inductance type soldered across the B+ and B- buss with the leads as short as possible. I used a double sided, copper clad circuit board and added 1/2 inch wide 0.062" thick bussing on the voltage and output rails to keep the vd/dt losses to a minimum. Proper spacing must be kept on the board traces to prevent flashover, as the buss voltage can exceed 200 volts with 140 volts AC input from the variac. For the input protection to the gates of the mosfets, I used two 10 volt / 1 watt zener diodes connected back-to-back. These diodes help protect the sensitive gates of the mosfets in the event of a stray input transient. The 5.1 ohm resistors with high speed diodes 1N5819 are conventional for input loading. I tried values from 1 ohm to 12 ohms here and found 5.1 ohms to be ideal. The gate driver transformers (GDT's) consist of dual core (each) with 15 turns on the primary and 16 turns (X2) on the secondaries. Using only a single core yielded poor waveform transfer under load, two cores were the best as three cores yielded no additional benefit. You may have to experiment with the number of secondary turns depending on the output transistors selected and frequency of operation. I actually used a different resonator than shown for my original experiments and found that the GDT as shown workes fine and only the R11 and R12 (3.9 ohms, 5 watt wirewound) resistors should be adjusted or "fine tuned" for optimal signal transfer. Expect to use a value of 2 to 6 ohms or so here. I used just enough resistance on R11 and R12 for good square wave output transfer with the input zener diodes getting only a little warm to the touch. Too much resistance causes a poor waveform resulting in poor switching (I squared R loss = heat), so everything is a balanced trade-off. R11 and R12 are mounted on the PLL / driver module and connected to the H-bridge power output module via small terminal blocks and a short twisted-pair jumper for each GDT input.
MOSFET's mounted on heatsink Top of circuit board. Note extra low loss buss.
Completed output PCB with GDT's Bottom of output PCB
Experimental GDT's Fully assembled power mosfet module
UC3710T DRIVER SECTION
The next step is to build the driver circuit. I noticed most experimenters liked using the 8 pin DIP gate driver chips such as UCC37221's. These 9 amp gate driver chips are suitable for driving the gates of most IGBT's and MOSFET's. What the manufacture doesn't tell you (unless you're really good at interpreting the data sheets) is that most of the time these chips are used in switching supplies that operate around 20 khz or so. So why is this important? Well, quite simply, most Tesla coils operate at several hundred kilohertz. This means that the gate needs to be switched on and off several hundred thousand time a second. The gate has a capacitive charge that needs to be overcome just to switch the mosfet on and off in the first place. This approximate capacitance can be found on the device's data sheet. For example the APT50M60JN power mosfet's I selected could have up to 14,000pf (14nf) of gate capacitance each, so if I'm driving two of these devices with one set of gate driver chips, the driver chips are loading into nearly 28,000 picofarads of load! Remembering that Xc is inversely proportional to the applied frequency, as the operating frequency goes up, the resistance goes down. So the higher the frequency your coil is designed to operate at, the more load is placed on the driver chips. The input capacitance of the APT50M60JN is so high, that the maximum allowed rating of the driver chips can easily be exceeded even at a low frequency. Much more discussion could be had here on the subject of driving a mosfet. I would recommend Steve Cloutier's excellent article on the subject for further study. If you intend to build a coil that operates above 175 khz, Chris recommends using a power mosfet with a much lower gate input capacitance such as the STE48NM50. These may be driven well beyond the maximum 175 khz of the APT50M60JN's with little or no modification to the UC3710T driver board assembly. The driver board assembly should draw no more than 700 mA total current from the DC power supply with all gates driven. If the current is higher than that, your operating frequency is too high. Remember, the higher the operating frequency, the higher the current required to drive the gates.
Since the driver chips on a mosfet Tesla coil operate under "brutal" conditions in the first place, I decided to follow Uzzor's lead. He uses the UC3710T driver chip on his audio modulated Tesla coil. This is a 5 pin device that looks a lot like a 7812 voltage regulator with the thermal tab mount. This is great because the device can be bolted directly to a heatsink to dissipate the additional heat caused by the higher operating frequencies encountered in Tesla coil service. You will notice that each driver pair has it's own 12 volt regulator with plenty of bypass capacitance placed as close as possible to the device(s). I'm using 47uf / 16volt tantalum bypass capacitors. When placed as close as possible to the driver chip's power rails, they aid in supplying the extra current needed during switching. It is important to note at this time that all chips should have at least a 0.1uf / 50 volt ceramic bybass capacitor placed as close as possible to the VCC to GND input of each device as well. All unused inputs on IC2 and IC3 should be grounded. Do not ground the outputs.
PLL OSCILLATOR SECTION
There are many PLL (Phase Locked Loop) IC's suitable for the oscillator. Many experimenters like to use either the popular TL494 or the 4046 PLL's for the main oscillator. Keep in mind that the 4046 comes in different flavors (5 volt and 15 volt models for example). I chose the 74HC4046 for my coil for a number of reasons. First of all, like the TL494 the 4046 can easily be FM modulated making it very easy to use. Secondly, there is only one adjustment required to tune the Tesla coil's resonant frequency. A bonus is that de-tuning this adjustment somewhat lower, the length of the spark can be increased. While making this adjustment, an oscilloscope should be connected to one of the mosfet outputs to observe that the signal stays clean. Too much adjustment in the wrong direction will cause all kinds of hash and trash to appear on the scope. This hash and trash can destroy the mosfet's in a big hurry. It is wise to play around with this adjustment and observe the input current meter and the wave form. After a bit of practice, you should be able to optimize and tune you coil "by ear". Steve Conner has an excellent article on tune up procedure for your coil. I recommend you follow it religiously.
CHOOSING R1, R2 and C1 values for the 4046 PLL
The math involved in working with PLL's can at the very least be quite daunting. So much so that entire college classes have been devoted to just their use. Thankfully Steve Conner made my life simple with his explanation of which components needed attention so I too would have success. A big thank you for that! I highly recommend referring to his article if you want to make your life easier.
Basically stated: R1 and C1 determine the frequency range of the internal VCO and R2 enables the VCO to have offset...that's it. Since I am using output mosfets with extremely high input capacitance, I was limited to a maximum usable frequency of about 175 khz to keep the input signal to the gates as clean as possible and keep from overloading the gate driver IC's. This presented a unique problem in that the secondary resonator needed to have a very low Fo. I chose a free running center frequency of about 140 khz or so. A value of 27k-ohms was chosen for R1 and C1's value was 300 pf. Placing R6 in the center of it's range gave me a frequency of about 140 khz. R2 (the offset resistor) is 39k-ohms. What the offset resistor does is give me a "window of operation" for the PLL's internal VCO of about 30 khz plus and minus from the center frequency of 140 khz. Anything above or below this window just shuts down the VCO and no output is available from the PLL. This is a very good and important feature because if the window were to be too wide, the PLL could possibly start up and operate the output stage and feed the primary coil with a second harmonic. This would lead to instant destruction of the Tesla coil and bridge! Recommendation: Do not permanently install R1, R2 and C1 until all testing is completed and the coil is up and running. The values of these components may need some tweeking to get it just right. Further information on R1, R2 and C1 including graphs and recommended values may be obtained by downloading Texas Instruments CD4046BE data sheet.
TESTING THE PLL
Once you have determined the center frequency, it is time to test and note the results. There are a number of methods available. I will describe what I did. Let's make the following assumptions: You have a resonator and top load that operates around 140 khz. You apply power to only the PLL and find that (with R6 at center slot) the PLL has an output of 125 khz. Change only the value of C1 from 300 pf so say 270 pf and this brings the PLL output up the 140 khz...perfect! But not so fast. You may discover later that the coil you bench tested at 140 khz actually operates better at 125 khz under full load (this happens quite frequently). Now you're back to changing C1 to 300 pf. That's why I say it is wise not to permanently install these critical components until you're ready to "box up" your project.
Now that you found the center operating frequency of the PLL, it's time to check the window of operation. With R7 removed from the circuit, inject a signal near the PLL's operating frequency into TP1. This is the feedback loop signal input path back to the PLL. You can tell that you have enough drive coming out of your signal generator because the PLL output frequency will "track" the input frequency as you vary the signal generator up and down from center slot. A this point the PLL is essentially "locked" on to the input frequency and will follow it until you hit the edge of the window. Let's say your tuning down and at 115 khz the PLL output suddenly jumps back to 140 khz, the lower edge of the window would be 115 khz. Now do the same for the upper edge and note the results. Let's say it's 160 khz. The window of operation would be 160 khz minus 115 khz or a 45 khz window. You should be able to see that there is no way the coil could start up outside this window.
SECONDARY FEEDBACK CT
You probably noticed CT-1 located on the earth side of the Tesla coil secondary. This is known as "secondary feedback". It works on the assumption that there is current at the base of the secondary. From the photograph you notice that the CT is cascaded. The effective ratio is about 1000:1. I wanted extremely fast voltage rise time so that the voltage feeding the PLL would be of proper phasing for turn on and shut down of the H-bridge. CT's need some sort of burden to work into, so D1 through D4 act as a voltage limiter clipping the signal when it exceeds about 6 volts or so. You should get a nice square wave output across the diodes when monitoring with a scope. R7 serves as a ballast resistor. You do not need to use this circuit if you don't want to. Just omit the CT, diodes D1 through D4 and R7 and connect a small four or five inch antenna to TP1 and place it near the secondary. Be sure to change C2 to a 1000volt or more ceramic or mica capacitor to protect pin 13 of the inverter if doing so. In fact I recommend not using the CT at all during tune up. The reason is the CT is polarity sensitive and if you have it hooked up backwards, you get absolutely nothing out of the secondary. However, once you have the Tesla coil optimized with only the antenna connected to TP1, the antenna can be removed and the CT installed and polarity checked. I got about 30% increase with the CT circuit as shown. Eliminating the antenna gives the Tesla coil a cleaner appearance.
BIPOLAR TESLA COILS
Bipolar secondaries have advantages and disadvantages over single-ended (monopolar) Tesla coils. The advantage is that you don't need a dedicated earth ground to operate a bipolar coil. The disadvantage is that you won't be able to use the feedback CT (CT-1) as shown, so you will need to use an antenna for the feedback loop connected to TP-1. I guess the toughest question here is: does the antenna go on the left or right side of the coil. Depends on the polarity of your primary. If it doesn't work, just reverse the primary leads, or move the antenna to the other side.
Another possible disadvantage of using a bipolar coil is by "just adding another top load" to the other end of a resonator causes the fundamental frequency of the resonator to increase by a factor of about 1.4 for the same coil were it to be monopolar. So for a conventional 1/4 wave resonator operating at say 140 khz, the expected bipolar operating frequency would be times 1.4 or about 196 khz. This means more work for the gate drivers due to Xc being inversely proportional to the frequency.
Insulation of the primary and secondary could be a challenge as well.
Photograph of OSC / PLL / driver board
Connector on the left is ANTENNA / CT-1 input, middle connector is output to GDT's, right is power in. Diodes CR3 (on the right) have been bypassed for DC power only. This was due to the voltage drop of the 12.6 volt / 2 amp filament transformer causing poor regulation. Note the single adjustment for tuning.
PRIMARY "TANK" CIRCUIT
Components "L" and "C" make up the tank circuit in a conventional coil. In a spark gap coil, for example, the values of L and C are chosen to be somewhere near and somewhat lower than the resonant frequency of the secondary and top load combined. A conventional Tesla coil also has a very low coefficient of coupling usually in the neighborhood of 0.12 to 0.18 or so.
The audio modulated coil differs substantially primarily because the coefficient of coupling is much greater than that of a spark gap coil. Wave damping becomes an issue. So the purpose of the MMC capacitor (C) is actually to block the DC and pass the AC to the primary, that's it. The primary is considered to be more of a transformer used to couple the energy from the H-bridge to the secondary. It is not a "tuned primary".
I used several 0.68uf / 630 VDC snubber style caps in a series / parallel combination for a total capacitance of about 1.3uf at 1.2kv. At full power, even after 30 minutes, they run ice cold.
The primary is 18 turns, 18" diameter spiral wound 1/2" spring tempered phosphorous bronze strap spaced 1/2" down to about 3 1/2" on the small end. Coupling is about 1/2" below the secondary.
Coupling should be adjusted by experiment before the coil is completely assembled. Once the resonator and top load are in place, the primary should be raised or lowered (if possible)-this was not possible in my case because the primary's small end is much smaller than the secondary. But if you make a primary so that can be slid up or down on the outside of the secondary, this should be an easy task. Just make sure you don't go too high on the secondary or you may flash over from the primary to the secondary and destroy the secondary. Primary can be of helix or pancake design. More turns may be required on a pancake due to inherent lower coefficient of coupling. Optimum coupling is when the coil can be run for several songs and the resonator gets "warm" to the touch. If the top load is too big, the audio may start sounding more distorted. Too large of a top load may also cause the secondary to overheat, so watch out for that. Too much input power will cause the secondary to overheat. Adjust the frequency fine tuning to reduce the power. If the audio is excessively distorted at a lower power level, reduce the coupling and increase the power. When everything is optimized, the coil should run somewhat warm to the touch and don't forget to check the waveform envelope across one of the mosfets to make sure the tuning is correct and the is no hash and trash on the waveform. All this takes time, so do it right the first time and you'll have a project that will yield hours of pleasure.
A little note about driving audio into the coil. The audio source can be from an I-POD, computer, stereo or other source. You may have noticed the small 5 inch bass speaker inside the cabinet. I use a small amplifier to boost the signal from the I-POD to an appropriate level to drive the speaker and coil to an appropriate level. Unfortunately the levels from most computers, etc. is usually a low level and not quite enough for full, rich sound. The Tesla speaker is a capacitive device which tends to produce only mids and highs, hence the use of the woofer to "fill in" the low end of the audio spectrum. An audio equalizer can be put to good use as well by splitting the highs into just the coil and lows into the speaker. Place the amplifier and audio source device as far away from the coil as practical. The coil emits a very large RF field as is evident by placing a fluorescent lamp in the vicinity. This field can "lock-up" computerized audio devices while the coil is in service. Have fun!
JANUARY 30th, 2011