Medium Sized Dual Resonant Solid State Tesla Coil
Medium Sized Dual Resonant Solid State Tesla Coil
Embedded Files
Build Process
Build Process
In the beginning of 2020, I was looking for a new project. I stumbled across Mads Barnkob's page detailing his DRSSTC, and I fell in love. I knew it was going to be a massive undertaking, but I felt I was up for the challenge.
In the beginning of 2020, I was looking for a new project. I stumbled across Mads Barnkob's page detailing his DRSSTC, and I fell in love. I knew it was going to be a massive undertaking, but I felt I was up for the challenge.
Before I begin, I want everyone to know that I take safety very seriously. Tesla coils are, with no doubt, extremely dangerous. The Pupman Safety Site has been useful.
Revision A - The First Shot
Revision A - The First Shot
The first step in making the coil was to wind the actual coil itself. I got a 6"x2' drain pipe from Lowe's and a spool of 30 AWG magnet wire to begin. Winding 22/0.0108 = 2037 turns of wire about the same size as human hair is very impractical and nearly impossible, so I set up a little winding machine to assist me. It started out as an extra stepper motor from my 3D printer held in place by a vise, but I have since made improvements detailed here.
The first step in making the coil was to wind the actual coil itself. I got a 6"x2' drain pipe from Lowe's and a spool of 30 AWG magnet wire to begin. Winding 22/0.0108 = 2037 turns of wire about the same size as human hair is very impractical and nearly impossible, so I set up a little winding machine to assist me. It started out as an extra stepper motor from my 3D printer held in place by a vise, but I have since made improvements detailed here.
This is a video of me winding the secondary coil. I used 8mm steel rods as the shaft on which the pipe spun, but the 3D printed joints I made for them broke when I was about halfway done.
This is a video of me winding the secondary coil. I used 8mm steel rods as the shaft on which the pipe spun, but the 3D printed joints I made for them broke when I was about halfway done.
DRSSTC secondary winding.MP4I eventually finished winding. I then coated the newly wound coil in varnish, which was definitely not the best choice. The initial layer I painted on was too thick and took many weeks to fully cure.
I eventually finished winding. I then coated the newly wound coil in varnish, which was definitely not the best choice. The initial layer I painted on was too thick and took many weeks to fully cure.
If at first you don't succeed...
If at first you don't succeed...
I've wound/acquired quite a few secondary coils over the years.
I've wound/acquired quite a few secondary coils over the years.
I went through a few toroid designs. I followed the path of many others, getting 4" aluminum duct and tying both ends together. I covered it in joint compound, which I let dry and then sanded down.
I went through a few toroid designs. I followed the path of many others, getting 4" aluminum duct and tying both ends together. I covered it in joint compound, which I let dry and then sanded down.
I used aluminum tape to cover the joint compound covered donut, but sadly the sticky side was not conductive, so I had to scrap this one. I eventually settled for just the plain duct donut without any "toppings."
I used aluminum tape to cover the joint compound covered donut, but sadly the sticky side was not conductive, so I had to scrap this one. I eventually settled for just the plain duct donut without any "toppings."
At the very top of the IGBT's datasheet, there is a line stating "for 20-60kHz switching." With the smaller toroid, the resonant frequency of the secondary was well above this. Using the bigger toroid increased the capacitance and lowered the frequency.
At the very top of the IGBT's datasheet, there is a line stating "for 20-60kHz switching." With the smaller toroid, the resonant frequency of the secondary was well above this. Using the bigger toroid increased the capacitance and lowered the frequency.
Before making the rest of the coil, I had to design a driver circuit which could drive the primary coil at its resonant frequency. I looked at Mads Barnkob's take on Steve Ward's driver, Steve Ward's universal driver, and the driver used in the oneTesla kit. I ended up going with a similar design to the oneTesla kit:
Before making the rest of the coil, I had to design a driver circuit which could drive the primary coil at its resonant frequency. I looked at Mads Barnkob's take on Steve Ward's driver, Steve Ward's universal driver, and the driver used in the oneTesla kit. I ended up going with a similar design to the oneTesla kit:
tesla sch v4.pdfThis circuit relies on a resistor network that allows the propagation of a signal from the interrupter without feedback from the primary coil. This network has proven to be finicky, so I plan to add more logic to discern between the signals in the future. The only difference between this schematic and my current board is the addition of an over current detection sub circuit, which is still in the works.
This circuit relies on a resistor network that allows the propagation of a signal from the interrupter without feedback from the primary coil. This network has proven to be finicky, so I plan to add more logic to discern between the signals in the future. The only difference between this schematic and my current board is the addition of an over current detection sub circuit, which is still in the works.
I used the SD interrupter from my smaller coil to signal this one. The maximum allowed on-time was only 50uS, which is not enough time for the larger coil's primary-current to ring up. Luckily, the developers made it straightforward to scale up durations in the source code, so I reprogramed the interrupter to allow up to 140us.
I used the SD interrupter from my smaller coil to signal this one. The maximum allowed on-time was only 50uS, which is not enough time for the larger coil's primary-current to ring up. Luckily, the developers made it straightforward to scale up durations in the source code, so I reprogramed the interrupter to allow up to 140us.
Up next is the board I designed. It is not the prettiest, but it works. (Future me here. Boy have I learned a LOT more about PCB layout since doing this design)
Up next is the board I designed. It is not the prettiest, but it works. (Future me here. Boy have I learned a LOT more about PCB layout since doing this design)
tesla v4.pdf
I ordered PCBs and got to assembling one. I mounted it inside of a 4" square box to act as a shield, as that is what I had on hand. There are still lots of openings and seams in this box, but it was hopefully better than nothing.
I ordered PCBs and got to assembling one. I mounted it inside of a 4" square box to act as a shield, as that is what I had on hand. There are still lots of openings and seams in this box, but it was hopefully better than nothing.
This is the schematic I used for the inverter. It's just your run of the mill H-bridge with a few extra components to help protect the transistors.
This is the schematic I used for the inverter. It's just your run of the mill H-bridge with a few extra components to help protect the transistors.
The only things I have modified are the values on some of the components. D1-D8 are TVS diodes, which are called to be 440v each, meaning a clamping voltage of 880v, much higher than the 650v rating on the IGBTs I currently have. I'm starting out with 2*250 for 500v and will increase when I am confident in the coil's function. I'm also changing D9-D16 to 20V Zeners to clamp the gate voltage at +-20V, which is the maximum gate voltage stated in the datasheet.
The only things I have modified are the values on some of the components. D1-D8 are TVS diodes, which are called to be 440v each, meaning a clamping voltage of 880v, much higher than the 650v rating on the IGBTs I currently have. I'm starting out with 2*250 for 500v and will increase when I am confident in the coil's function. I'm also changing D9-D16 to 20V Zeners to clamp the gate voltage at +-20V, which is the maximum gate voltage stated in the datasheet.
These are all the parts I ordered. Most notable are the insulated-gate-bipolar-transistor bricks, tank capacitors, and massive 7,500uf 450v electrolytic bus capacitors.
These are all the parts I ordered. Most notable are the insulated-gate-bipolar-transistor bricks, tank capacitors, and massive 7,500uf 450v electrolytic bus capacitors.
Once I was done gawking at everything, I got to constructing the high-power side of the circuit. I made a rudimentary base from some wood and dowels and started laying out the components. This configuration turned out to be one of the worst ways I could have set it up. Voltage transients can be calculated from V=L*dI/dt, and as high currents are being switched at high frequencies, it is vital that inductance is kept to a minimum. This first layout has lots of "loop area" between the IGBTs on the left and the capacitor bank on the right, meaning lots of extra inductance is present which can and will produce voltage spikes that could fry expensive components. While there are TVS diodes and snubber capacitors protecting the IGBTs, it is bad practice to make a design that functions poorly which relies on protective features.
Once I was done gawking at everything, I got to constructing the high-power side of the circuit. I made a rudimentary base from some wood and dowels and started laying out the components. This configuration turned out to be one of the worst ways I could have set it up. Voltage transients can be calculated from V=L*dI/dt, and as high currents are being switched at high frequencies, it is vital that inductance is kept to a minimum. This first layout has lots of "loop area" between the IGBTs on the left and the capacitor bank on the right, meaning lots of extra inductance is present which can and will produce voltage spikes that could fry expensive components. While there are TVS diodes and snubber capacitors protecting the IGBTs, it is bad practice to make a design that functions poorly which relies on protective features.
After some revision, this is bridge design I made for my first successful test. I really could do some more work on it due to the reasons above, and given the fact that I have now blown IGBTs twice.
After some revision, this is bridge design I made for my first successful test. I really could do some more work on it due to the reasons above, and given the fact that I have now blown IGBTs twice.
Below is just a sketch of what a future bridge layout might look like. I plan to use cut copper sheets to form busbar which will bring inductance to an absolute minimum. The four rectangles on top represent the IGBTs, and the bars are connected directly to the capacitor bank.
Below is just a sketch of what a future bridge layout might look like. I plan to use cut copper sheets to form busbar which will bring inductance to an absolute minimum. The four rectangles on top represent the IGBTs, and the bars are connected directly to the capacitor bank.
This is the MMC bank, short for multi-mini-capacitor. It is the tank capacitor for the LC circuit in the primary coil. I put six parallel strings of two caps in series (each 0.15uf @ 2000VDC), giving an equivalent rating of 4000 VDC at 0.450uf of the renowned 942c series by CDE. These capacitors have a dV/dt of nearly 3000 V/us, which I honestly didn't even know was possible until I saw the datasheet. They are also used for their high peak currents and self-healing abilities. The two connectors on each side go to 6 AWG stranded wire.
This is the MMC bank, short for multi-mini-capacitor. It is the tank capacitor for the LC circuit in the primary coil. I put six parallel strings of two caps in series (each 0.15uf @ 2000VDC), giving an equivalent rating of 4000 VDC at 0.450uf of the renowned 942c series by CDE. These capacitors have a dV/dt of nearly 3000 V/us, which I honestly didn't even know was possible until I saw the datasheet. They are also used for their high peak currents and self-healing abilities. The two connectors on each side go to 6 AWG stranded wire.
I went through multiple attempts at making a decent primary coil. This was my first stab at a flat spiral configuration, and to be frank, it looked atrocious. Needless to say, it did not make the final cut.
I went through multiple attempts at making a decent primary coil. This was my first stab at a flat spiral configuration, and to be frank, it looked atrocious. Needless to say, it did not make the final cut.
One of my later attempts looked much better. I used a bucket to form 3/8" OD copper tubing into a nice cylindric shape. I was afraid, though, the top was too close to the toroid on the secondary, so I went back to a pancake spiral--->
One of my later attempts looked much better. I used a bucket to form 3/8" OD copper tubing into a nice cylindric shape. I was afraid, though, the top was too close to the toroid on the secondary, so I went back to a pancake spiral--->
I settled with this spiral I made. To save copper, I wound it as tight as physically possible without it touching itself. I used zip ties at spots where I felt needed support. Zip ties may not be the best option, as I am not sure yet of how hot the primary may get, and the ties are made of plastic.
I settled with this spiral I made. To save copper, I wound it as tight as physically possible without it touching itself. I used zip ties at spots where I felt needed support. Zip ties may not be the best option, as I am not sure yet of how hot the primary may get, and the ties are made of plastic.
With the primary done, I had to tune its inductance to have a resonant frequency slightly less than that of the secondary (a spark adds capacitance to the secondary, reducing its resonance. The resonant frequency of any inductor-capacitor set up is inversely proportional to the square root of the inductance times the capacitance). This was easy enough with the help of JavaTC, a program that runs all sorts of calculations for Tesla coils. I got an idea of how many turns the primary should be, and then fine tuned it with an oscilloscope and signal generator (which were so generously borrowed from a retired electrical engineer who has been a mentor to me). With that all set, the next step was to test it out! After a full system check, I brought the beast out of my basement into my garage. I put on my goggles and ear protection, cranked the variac, and nothing. Nothing happened. I was half expecting nothing to happen; rarely anything works on the first try. The coil won the battle but I was determined to win the war!
With the primary done, I had to tune its inductance to have a resonant frequency slightly less than that of the secondary (a spark adds capacitance to the secondary, reducing its resonance. The resonant frequency of any inductor-capacitor set up is inversely proportional to the square root of the inductance times the capacitance). This was easy enough with the help of JavaTC, a program that runs all sorts of calculations for Tesla coils. I got an idea of how many turns the primary should be, and then fine tuned it with an oscilloscope and signal generator (which were so generously borrowed from a retired electrical engineer who has been a mentor to me). With that all set, the next step was to test it out! After a full system check, I brought the beast out of my basement into my garage. I put on my goggles and ear protection, cranked the variac, and nothing. Nothing happened. I was half expecting nothing to happen; rarely anything works on the first try. The coil won the battle but I was determined to win the war!
After lots of late nights scratching my head, I narrowed the problem down to the gate drive transformer. The gate drive transformer separates the logic components from the intimidating primary circuit. Without doing too much research, I followed Mads and took cat5 networking cable and wound it around a high permeability toroidal ferrite core. Driving the transformer with a dummy 70kHz feedback signal from the signal generator and measuring the output with the scope gave this picture:
After lots of late nights scratching my head, I narrowed the problem down to the gate drive transformer. The gate drive transformer separates the logic components from the intimidating primary circuit. Without doing too much research, I followed Mads and took cat5 networking cable and wound it around a high permeability toroidal ferrite core. Driving the transformer with a dummy 70kHz feedback signal from the signal generator and measuring the output with the scope gave this picture:
The two traces are two of four secondaries coming from the transformer. The two signals are inverted--as they should be. The problem lay with me getting too excited when I was nearing finishing the bridge. I accidentally swapped two of the secondaries so that IGBTs 1-3 would be on at one time and IGBTs 2-4 the next (refer to schematic above). Thankfully, due to sheer luck, IGBTs 1-2 and 3-4 were not paired, as they would have shorted the capacitor bank and violently blown. I fixed the problem, double and triple checked that I had actually fixed it, and tested it as soon as I could.
The two traces are two of four secondaries coming from the transformer. The two signals are inverted--as they should be. The problem lay with me getting too excited when I was nearing finishing the bridge. I accidentally swapped two of the secondaries so that IGBTs 1-3 would be on at one time and IGBTs 2-4 the next (refer to schematic above). Thankfully, due to sheer luck, IGBTs 1-2 and 3-4 were not paired, as they would have shorted the capacitor bank and violently blown. I fixed the problem, double and triple checked that I had actually fixed it, and tested it as soon as I could.
Over to the right showcases my first successful test! Sadly, 2 of the IGBTs blew right after I shot the video. I have since fixed the issue and improved the design. See below.
Over to the right showcases my first successful test! Sadly, 2 of the IGBTs blew right after I shot the video. I have since fixed the issue and improved the design. See below.
Theory of Operation
Theory of Operation
A Tesla coil is an air cored transformer with some very unique properties. If the two coils did not resonate, one could expect 200v*(2200/4.5)~=100kv on the output. However, this number will be lower as the coupling between the primary and secondary is pretty low. Instead, the two coils resonate at the same frequency. Resonance is a special property in physics, occurring when a system is oscillating at the frequency it naturally wants to. In a dual resonant Tesla coil, the two coils consist of a coil (inductor) and a capacitor (for the secondary, the toroid is the main source for capacitance to the world). The oneTesla tutorial page describes this phenomenon very well, too well for me to not include it on my page:
A Tesla coil is an air cored transformer with some very unique properties. If the two coils did not resonate, one could expect 200v*(2200/4.5)~=100kv on the output. However, this number will be lower as the coupling between the primary and secondary is pretty low. Instead, the two coils resonate at the same frequency. Resonance is a special property in physics, occurring when a system is oscillating at the frequency it naturally wants to. In a dual resonant Tesla coil, the two coils consist of a coil (inductor) and a capacitor (for the secondary, the toroid is the main source for capacitance to the world). The oneTesla tutorial page describes this phenomenon very well, too well for me to not include it on my page:
"Consider what happens when you don’t drive the circuit (assume that the AC source in the figure is replaced by a wire), but start out with the capacitor charged. The capacitor wants to discharge, so charge flows around the circuit, through the inductor, to the other plate. In the process, a magnetic field builds up inside the inductor. When the charge on each plate of the capacitor is zero, current stops flowing. But at this point, the inductor has energy stored up in a magnetic field - which tends to oppose change. The magnetic field collapses, inducing a continuing current in the same direction, thereby recharging the capacitor and restarting the cycle in the opposite direction."
A good parallel to resonance is pushing someone on a swing. By pushing every time the person is swinging away from you, the person on the swing can reach great heights. In the same way, by carefully timing the addition of energy to the system in sync with these oscillations, massive peaks are achieved. The secondary coil (ideally) resonates at the same frequency as the primary, leading to ludicrous outputs. There are many other factors in operation, but they are outside of the scope of my build description.
A good parallel to resonance is pushing someone on a swing. By pushing every time the person is swinging away from you, the person on the swing can reach great heights. In the same way, by carefully timing the addition of energy to the system in sync with these oscillations, massive peaks are achieved. The secondary coil (ideally) resonates at the same frequency as the primary, leading to ludicrous outputs. There are many other factors in operation, but they are outside of the scope of my build description.
Revision B - Hindsight is 2020
Revision B - Hindsight is 2020
As stated previously, my original bridge design had lots of stray inductance. Below are some of the steps I took to fix that issue.
As stated previously, my original bridge design had lots of stray inductance. Below are some of the steps I took to fix that issue.
I drew my sketches of the new bridge layout in AutoCad and got them commercially laser cut in copper. I've been wanting to experiment with laser cutting, and I finally found an excuse. Here are the parts:
I drew my sketches of the new bridge layout in AutoCad and got them commercially laser cut in copper. I've been wanting to experiment with laser cutting, and I finally found an excuse. Here are the parts:
This is them on top of the sketch I drew. It's really nice when things work out the first try. Measure-twice-cut-once is a virtue!
This is them on top of the sketch I drew. It's really nice when things work out the first try. Measure-twice-cut-once is a virtue!
I also bought a single large heat sink to mount all of the IGBTs on together. It looks nicer, is easier to use with the busbar, and allows for more airflow.
I also bought a single large heat sink to mount all of the IGBTs on together. It looks nicer, is easier to use with the busbar, and allows for more airflow.
I also went ahead and upgraded the IGBTs. The original 72n60's I had were simply not designed for 80kHz operation.
I also went ahead and upgraded the IGBTs. The original 72n60's I had were simply not designed for 80kHz operation.
Here it is, set up and in place. I decided to mount the heat sink backwards, as I was not sure how hot it would get, and it was close to the capacitor bank.
Here it is, set up and in place. I decided to mount the heat sink backwards, as I was not sure how hot it would get, and it was close to the capacitor bank.
In addition to the new bridge, I decided to buff up the MMC. The new setup uses 3x7 caps for 0.35uF @6kVDC, or ~850Vrms derated at a resonant frequency near 80kHz. Adding more parallel strings also increases the Irms rating.
In addition to the new bridge, I decided to buff up the MMC. The new setup uses 3x7 caps for 0.35uF @6kVDC, or ~850Vrms derated at a resonant frequency near 80kHz. Adding more parallel strings also increases the Irms rating.
It's double sided, with four strings on top and three on the back. There is also room to add another string if I ever need to.
It's double sided, with four strings on top and three on the back. There is also room to add another string if I ever need to.
Success! Or at least somewhat. A key feature in this new design was the addition of an overcurrent detection circuit (OCD), which pulls the interrupter signal low if the current in the primary gets too high. This is done with another 33:33:1 cascaded transformer, which will roughly output 1/1000th of the primary current. The output is burdened by a 5 ohm resistor, which then is fed into a comparator. The comparator, when triggered, triggers a 555 timer in a monostable configuration which then triggers a MOSFET which pulls the interrupter line low. I tested the coil with OCD set to 200A pk, and I got sparks so tiny I could only hear them, not see them. I tried raising the threshold significantly on the comparator and got the same results, so I decided to try removing the OCD feedback altogether and it worked! I went back later on and tried the OCD again and it functioned properly. I think a wire may have been brushing where it should not have been.
Success! Or at least somewhat. A key feature in this new design was the addition of an overcurrent detection circuit (OCD), which pulls the interrupter signal low if the current in the primary gets too high. This is done with another 33:33:1 cascaded transformer, which will roughly output 1/1000th of the primary current. The output is burdened by a 5 ohm resistor, which then is fed into a comparator. The comparator, when triggered, triggers a 555 timer in a monostable configuration which then triggers a MOSFET which pulls the interrupter line low. I tested the coil with OCD set to 200A pk, and I got sparks so tiny I could only hear them, not see them. I tried raising the threshold significantly on the comparator and got the same results, so I decided to try removing the OCD feedback altogether and it worked! I went back later on and tried the OCD again and it functioned properly. I think a wire may have been brushing where it should not have been.
This is an image from the above video. Based on the fact that the toroid has a major-diameter of 18", this spark is longer than three feet, shockingly at less than 1/2 max power!
This is an image from the above video. Based on the fact that the toroid has a major-diameter of 18", this spark is longer than three feet, shockingly at less than 1/2 max power!
11 months of work (in addition to being a full time student, athlete, and having a job) have paid off in a good way so far. However, there is always more to do! It turns out the OCD was brushing up against something it wasn't supposed to. I caught that, and everything is working now.
11 months of work (in addition to being a full time student, athlete, and having a job) have paid off in a good way so far. However, there is always more to do! It turns out the OCD was brushing up against something it wasn't supposed to. I caught that, and everything is working now.
Revision C - New Driver
Revision C - New Driver
A natural consequence of having a bigger 'n badder primary circuit is having bigger 'n badder fails. The coil blew during a run and left a pretty big mark. Below is some of the busbar covered in the soot created during IGBT failure. Sadly, I somehow managed not to capture the fail on video. It appears to have gotten so hot that the M4 bolt securing the IGBT was welded to the nut.
A natural consequence of having a bigger 'n badder primary circuit is having bigger 'n badder fails. The coil blew during a run and left a pretty big mark. Below is some of the busbar covered in the soot created during IGBT failure. Sadly, I somehow managed not to capture the fail on video. It appears to have gotten so hot that the M4 bolt securing the IGBT was welded to the nut.
What does this mean? An excuse to upgrade it, of course! The driver I designed definitely can use some improvements. I found Loneocean's UD2.7C and got intrigued. His board offered a higher gate-drive voltage by virtue of using MOSFETs instead of gate drive ICs and also had phase-lead, which helps eliminate switching spikes. I figured, I'm already this far into the build, I've had my shot at a driver board, it's time to upgrade. His board has been tried and tested by many and is proven to work for coils in all shapes and sizes. Below is a little video in action. The sparks could definitely be bigger - I just have to fine tune the resonance of the primary.
What does this mean? An excuse to upgrade it, of course! The driver I designed definitely can use some improvements. I found Loneocean's UD2.7C and got intrigued. His board offered a higher gate-drive voltage by virtue of using MOSFETs instead of gate drive ICs and also had phase-lead, which helps eliminate switching spikes. I figured, I'm already this far into the build, I've had my shot at a driver board, it's time to upgrade. His board has been tried and tested by many and is proven to work for coils in all shapes and sizes. Below is a little video in action. The sparks could definitely be bigger - I just have to fine tune the resonance of the primary.
Sneak peak of what else I'm working on. I asked myself, "What's cooler than a Tesla coil?" The obvious answer is two Tesla coils! I am currently heads down in the interrupter's software. Stay tuned!
Sneak peak of what else I'm working on. I asked myself, "What's cooler than a Tesla coil?" The obvious answer is two Tesla coils! I am currently heads down in the interrupter's software. Stay tuned!
Page updated
Google Sites
Report abuse