The QCW works on a very similar mode of operation to the VTTC. The system power supply is a ramp, or half sine (in the case of a VTTC), and this is fed to a high Z high fres Tesla coil usually in the order of 20 ohms+ and 300KHz+. This coupled with the slow ramp-up time in the mS range, gives the sword-like appearance of the sparks.

The QCW can be broken into 3 main parts.
  • Modulator
  • DR driver 
  • High Z Tesla Coil resonator 


The modulator is one of the most important parts of this entire build. My modulator is a class D delta modulator. The class D amplifier is a half or full bridge feeding a low pass LC filter. The bridge is driven with a PWM signal that chops up it's 380VDC bus and sends that through its low pass LC filter. The filter gets rid of all the high frequency switching hash and leaves only the average voltage of the PWM. The output LC has a low impedance, making it's slew rate rather fast, giving the converter the ability to change it's output voltage level in a matter of uS. 

The control method I use to drive the class D is known as delta modulation. Delta modulation is a much better control method than a proportional control loop. This is due to its stability under almost any condition - unlike a proportional controller that can oscillate out of control or have low bandwidth if not tuned right.  Delta modulation is also referred to as Hysteresis Control or Bang Bang control.  

The control loop has an integrating op amp that integrates the error signal before feeding it to a comparator -  in order to square it and then turn it into a suitable signal for gate drive. At low power the switching frequency will be low, but as power increases the switching will speed up to compensate and add more power to the output. The faster the op amp integrates the signal, the faster the overall switching speed will be, so the integration speed must be set correctly to suit the output filter and switches used for the half bridge.

The output LC is another part of the system that needs special attention. Keep in mind that because there is no bulk capacitance for the main inverter to pull from the peak current, the current will be in the tank LC of the Tesla coil which will also be pulled from the modulator. This current can be as high as 250Apk in extreme cases. So the half bridge switches and inductor have to be rated to handle the peak tank current. My QCW doesn't run any higher than 120Apk, so the output inductor of the class D had to not saturate at 120Apk or less. It also had to have an inductance of 250uH to match up with the 10uF C to make a low pass filter with a pole at 3KHz. (switching freq ~20KHz). For this I used a stacked ferrite E core core with a 3/8 inch air gap.

This is the modulator control circuitry. JP2 is a transformer input and this input must be isolated!!! The inputs are from a single transformer with isolated secondary windings. JP6 is for a LEM100 DC current transducer used for cycle by cycle limiting, used primarily if there is a load fault on the output (aka blown up bridge). The OCD circuit was never implemented in my final design and needs some revising and tweaking to work properly. The experienced eye will be able to see this and work around these problems. The HFBR-2412T receives a PWM signal over fiber that is converted to an analog wave shape by L1 and C2. HFBR-1412T is used to send an enable interrupter signal to the DRSSTC driver board. Also C4 is used to set the intonation rate of the system. JP3 is feedback for the system from a HV resistor divider on the output of the converter.        

The modulator basically takes the 380VDC bus that we create for it from a large storage cap (this should be a big cap, say 10mF) and converts it to a signal of our choice that can swing from 0 to 380VDC. The signal that works best for long strait sparks in the QCW, is a long linear ramp 5mS to 20mS long with a 2mS to 5mS ramp down time. The ramp down is there to bring the voltage back down slowly after a burst so that you don't get a large "pop" in the output. I usually bring the voltage back down to 30VDC and float it there so that the DRSSTC drive can start up oscillating correctly at the beginning of a burst. It's very important that there is not a large voltage sitting on the inverter input when it starts up as this can blow-up the inverter.   



The left photo is on a resistive test load, yellow is the inverter output, and the blue is the half bridge switching waveform. Notice how the frequency gets higher with higher output power. Right photo is the QCW in operation, yellow is inverter output notice how it's floating at 40VDC and blue is primary current about 28Apk in this case ( that's about 1 foot of sparks maybe 2).


On the left is the class D delta modulator driving a resistive test load. On the right is the delta modulator wired into the system in a lash up setup.

High Z Tesla Coil Resonator

The most challenging part was getting the modulator working. It's a difficult SMPS to get working. A significant thing that sets the QCW apart is its high Z high-frequency tank. If you have ever done a VTTC this should be nothing new. The idea with the QCW is to have a high impedance tank circuit that helps keep the primary current low; this is essential because the pulse length is so long 25mS in some cases. The high frequency is important because it's required to keep branching to a minimum. Fres <300KHz tend to branch more while >300KHz tend to stay straight, getting straighter as the freq increases. 

The primary configuration for my system was based on the tank capacitor I had on hand. The tank cap I used was a 10nF 10Kv MICA given to me by Dr.Spark (Thanks Dr. Spark!) Unlike a regular DRSSTC, the peak current is not all that high, and as a result, the tank voltages are also rather low. The only rating that has to be carefully looked at is the RMS current handling rating of the capacitor. This is very important because of the high duty cycles the QCW runs at. 

I've tried many different types of tank capacitor in my QCW and I will give the results of my experimentation with them:

The Teflon -> They work and they work HOT!!! Not the best...but they do work... sort of.

Poly -> Although they are great for high current pulsed applications like a regular DR, they tend to get hot in the high RMS environment of the QCW. If you made an MMC you would need many caps in parallel for a good RMS rating. The plus side is that because the peak current is low, the voltage rating can also be very low - probably only two or three 2kv caps in series for most applications. 

MICA -> They're awesome - most mica caps can have great RMS ratings and already have a nice high voltage rating. I use a single 10nF 10Kv MICA rated at 18Arms in my QCW and it doesn't break a sweat at 20pps 100Apk in the tank. I highly suggest MICA for the QCW, if you can find them!  

The primary was designed for high coupling with the secondary. Because this system has very low output voltage (<70Kv in most cases), the primary and secondary can be coupled very tightly without flash over. This results in two things: the upper pole of the secondary is moved way up and the energy transfer is very fast. The coupling of my system is 0.369K

The secondary is a 4.5 x 9 inch winding of 30 AWG wire with a 2 x 8 topload. The fres of the secondary circuit is about 280KHz, but with the high coupling, this puts the upper pole at about 320KHz. This is the perfect frequency for the QCW as it will result in nice strait sparks.  As always, experimental tuning is always the best way to find the optimal tuning point, although the QCW needs a little extra math beforehand. The optimal mode of operation for the QCW has yet to be confirmed, so lower pole or bang on resonance tuning are certainly things to try in the future.   

One other important note for the secondary circuit is the use of a good breakout point. Small wire will quickly melt away leaving molten blobs of wire on your toroid! Thick steel or tungsten should be used if possible because those sparks are hot!

DR Driver 

For the most part, this is just a regular DRSSTC driver with primary current feedback and a full or half bridge of IGBTs switching at the zero crossings. There are, however a couple things to take note of that impact this significantly.

  • Use of fast IGBTs like the IXYX 60N60s is a must! Because they will be switching at high frequencies (300KHz+) and also because they are being drive for so much longer, the switching losses will be much higher. Unlike a regular DR that runs 4 to 10 cycles a burst the QCW, runs thousands of cycles a burst!!!
  • Use of a phase lead is highly recommended for the same reasons as using fast IGBTs.
  • Water cooling or very good forced air cooling setup is also needed. Because the IGBTs will be running high RMS currents, it's important that they're kept cool. Cool IGBT generally switch faster and have lower conduction losses as well. 
  • Keep the snubbed C on the IGBTs low <5uF. If this is too high it can can mess with the regulation of modulator and the performance of the inverter. The one exception to this is when the modulator output LC capacitor is mounted directly to the bridge. 
  • Inductance between modulator output and the inverter should also be kept at a minimum.   

  I also found that a good gate drive setup is required for reliable operation. I used a large GDT to drive my full bridge of 60N60s. 

On the left is the gen one water cooled 60N60 full bridge. On the right is a lashed up test setup.

Future Updates 

This project is still a work in progress and will be updated as work progress. Some things that I will be trying and/or implementing in the future:

  • Larger IGBT switches for bigger sparks!
  • Higher freq secondary in the 450KHz+ range
  • Audio modulation of the bus voltage using class D delta modulator
  • Bigger sparks!

I want to give a big thanks to all the people that have helped me along the way with this project:

More videos can be found on our YouTube Channel 
More Photos can be found on our Flikr page