DRSSTC 3 Twins


Topload 18"x4.5"
Secondary coil 4.5"x15.5" 2000 turns 34AWG
Primary coil 9 turns of 1/4" copper tubing
Tank capacitor 0.15uF 8kV, 4 series 4 parallel strings of 942C20P15K-F; later added another string for 0.1875uF
Resonant frequency 93kHz, tuned to 83kHz
Peak primary current 522A (OCD)
Bus capacitor 2x 3300uF 400VDC on one coil and 2x 4700uF 450VDC on the other
Bus voltage 370-670VDC
Bridge CM200DU-24F full-bridge (RTC removed)
Controller UD 2.5 REV A
Maximum spark length 50"

Introduction

Upon the aquisition of a number of CM200DU-24F bricks (from a Toyota Prius?), I decided to make some large Solid State Tesla Coils... because why not? :) These coils later became the subject of my high school physics project. (January 2014 - June 2014)

Secondary Coil and Toroid

The secondary was wound with 2000 turns of 34AWG magnet wire over 15.6" on a 4.5" diameter PVC pipe. 32AWG wire probably would have been a better choice, but that would extend the winding length to the point where the coil looked awkwardly tall. The pipe was prepared by scraping off the black markings with a knife and light sanding. Two end caps were made from 3/4" wood and they each have a 1/4" bolt through the middle, were secured by three nylon screws. Hot glue was dribbled over the end of the screw on the end cap to provide some insulation. The end of the magnet wire was soldered to a short length of bare 26AWG wire and fed through a small hole on the side of the end cap that went to the middle where the bolt was. The wire was inserted and the bolt was screwed in, holding the wire and making contact. The coil was coated in four layers of Minwax Fast Drying Polyurethane (took a few days to dry...), and the ends were held down with packing tape. Most of the resonator calculations were done on DeepFriedNeon.

The toroid was made out of 4.5" diameter dryer duct tubing and some 1"x1/8" aluminum bar. The aluminum was bent into a circle, and the dryer duct tubing was bent around it and secured with some magnet wire and a few screws. Total cost: $10. Spun aluminum toroid: $229.

Primary Coil and Tank Capacitor

50' of 1/4" copper refrigerator tubing from eBay took care of the two primary coils. I was going to buy some 3/8" copper tubing for the strike rails, but knowing that there was going to be a little extra left over, I just used the 1/4" tubing for the strike rail. Ideally a large diameter strike rail would reduce the chance of corona breakouts, but this coil didn't seem to have any issues with that. The 1/4" thick acrylic primary supports were designed on AutoCAD and cut out on a bandsaw. Surprisingly, all but three notches were a snug fit. The supports were drilled and tapped to be mounted to the plywood base with nylon fasteners. The primary circuit was wired together with 6AWG wire. I found out later that 800Apk heats the copper tubing enough to melt some of the acrylic supports...

Each tank capacitor is made of 16 942C20P15K-F capactiors, giving a rating of 8kV, 0.15uF. Resonance voltage rise brings the primary voltage up to about 7kV. The capacitors are connected in parallel with some braided copper wire (leftover from another project) and put in an acrylic case designed in AutoCAD. Holes probably should have been drilled on the side to help dissipate heat, but I couldn't be bothered, and I would also risk cracking the acrylic. Calculations showed that the capacitors would have a minimal temperature rise during full operation, and it held to be true in reality. MMC calculations were done on Mads Barnkob's website.

H-Bridge Inverter

The IGBT's used were rated for 1200V, 200A continuous (400A pulsed). The coil was designed to run on 240VAC input (670VDC bus) to take advantage of the voltage rating. Primary current was designed to be around 600A, but the F series of the CM modules have a built-in RTC circuit that shuts off the IGBT if the current passes 400A. The IGBT was opened, and the wires to the circuit were removed. At first, I used a pair of scissors to cut the bond wires, but I found it easier to just take some sharp tweezers and wiggle the wire until it broke off. It was also much cleaner and more of the thermal goop stayed intact. The face of the aluminum heatsinks had a slight concave bend, so there would be a gap between the back of the IGBT and the heatsink when mounted. I flattened it with a belt sander and some fine sandpaper. Thermal paste was not used :)

So many bus layouts... which do I choose? After a while of fumbling around with huge electrolytic capacitors and IGBT bricks, I decided on the bus layout shown below (EAGLE CAD screenshot). The bus capacitors are mounted only by the aluminum bus bars, and they're quite sturdy!

The bus work was done with 1/4" aluminum bars. A laminated bus design was considered, but the aluminum I had was rather thick, and it would have used much more material. A 1000V, 50A bridge rectifier was used because somehow it was cheaper than the 30A bridge rectifiers, but who knows if they are up to spec? Anyway, I wouldn't have to worry about it dying anytime soon. A 2000V, 0.10uF capacitor was put between the negative rail and ground to absorb some of the spark's current if it hit the primary coil (Steve Ward method). Two 1000V, 2uF snubber capacitors absorb some of the voltage spikes at each IGBT. Each IGBT gate is protected by a 33V transient voltage supressor. The system is grounded according to Steve Ward's method.

Driver

The coil is controlled by the Universal DRSSTC Driver Version 2.5. Seeedstudio is great; I ordered 10 boards and got 11, and the quality is very nice.

The two boards took about four hours to assemble, which isn't bad considering that I have never done SMD soldering before. The boards went together with no issue, apart from one of the fiber optic receivers (IF-D95T) not working; perhaps I just got a bad batch of them. The drivers reside in two aluminum project boxes acquired from RadioShack. The phase lead inductors were obtained from Coilcraft.

The gate drive transformers were made of 2" diameter ferrite cores (they are huge!) and CAT-5e cable. I fit as many turns as I could on it, which turned out to be 14, to prevent the IGBT gate from saturating. Ferrite beads were added to the primary side of the gate drive transformer to reduce the amount of noise that gets into the driver.

Testing

Initial testing was relatively straightforward. The UD 2.5 was tested by scoping the output of the gate drivers whilst applying a feedback signal via a signal generator. Next, the gate drive transformer was connected, and the bridge was powered with a variac. The circuit began oscillating at around 15VDC bus with feedback from the current transformers (didn't bother with the signal generator).

It took about 30VDC before I could hear the primary vibrating. The primary currrent was scoped by another current transformer, and everything looked fine on the scope, apart from the far-from-square-wave looking bridge output. Still not sure why, scope probe interference? The bridge output was scoped by using a probe on each output and adding channel A to the inverse of channel B. Note the 50W 10k power resistor across the bridge output. The resistor discharges the tank capacitor after each pulse, so there is no voltage offset at the beginning of a cycle leftover from the previous cycle.

The photo at the top is from the first coil's first light. The coil happily made 4' sparks on 800Apk primary current. The only issue was flashovers that occurred between the primary and secondary. On a later run, flashover ended up burning a hole in the secondary, severing four turns. The burnt wires were removed, and the carbon was scraped off the PVC pipe. The ends were soldered together, and the secondary worked again! To stop the flashovers, the secondary was installed upside down, effectively raising it since the secondary was wound with 3/4" of space on one end and 2" on the other.

The second coil was finished a few weeks later (I was waiting for more bus capacitors and a fiber optic receiver). Like the first coil, it worked without any issues, and the secondary was flipped to prevent flashovers. The coil was run on full power, 240VAC input, 670VDC bus :)

The best settings for taking pictures of the coil are low ISO, low light, and a shutter speed of a few seconds.

Update 2-14-15

I decided to take some measurements on the resonant frequencies and found that the primary was tuned to 90kHz, only 3kHz below the secondary (upper pole). Adding a 4 foot wire to the secondary dropped the resonant frequency to 80kHz. There weren't any more turns on the primary coil, so the easiest solution was to add another string of capacitors to the MMC. The OCD was originally set at 880A since I just kept turning it up until the coil made sparks...

The coil now runs with 25% less pulsewidth than before, and the maximum arc length increased by a foot! Sparks randomly break out from the toroid occasionally, and there are a few flashovers at low break rates. The coil now runs on 800A.

Update 11-16-16

Still going strong with no issues!

Update 4-21-17

MMC capacitor bank on coil #2 died. Noticed small spark output with the OCD light on, and then realized there were sparks coming from under the coil. Upon further inspection it seemed that it was arcing between capacitors in the MMC, but after it was disassembled, two out of twenty capacitors measured nothing on a capacitance meter. I have some spare caps lying around, so hopefully replacing those two will fix it. Unsure if this is a one-off failure, or if the other capacitors will die soon as well. MMC only felt warm to the touch after extracting it from the coil.

MMC was repaired, two caps were replaced , and the coil was working again.

Update 7-1-18

Ordered new phase lead inductors and swapped the old ones out. Had 7M3-153 and 7M3-223, replaced both with 7M3-393. Blue waveform is one output of the bridge, yellow is primary current, and purple is gate voltage.

Before:

After:

Update 8-27-17

Coil #2's MMC dies again. Haven't gotten a chance to see exactly which and how many caps died. I'll be ditching this MMC design as it seems to be slightly underrated in voltage. Youtube link: